[llvm] r314729 - Re-enable "[MachineCopyPropagation] Extend pass to do COPY source forwarding"

Geoff Berry via llvm-commits llvm-commits at lists.llvm.org
Tue Oct 3 09:05:14 PDT 2017


Hi Quentin,

Yes, I'll revert shortly.  The verifier issue sounds like an issue I 
though I had worked around regarding forwarding to undef uses.  Let me 
know if I can help.

-Geoff

On 10/3/2017 12:01 PM, Quentin Colombet wrote:
> Hi Geoff,
> 
> I see the MachineVerifier complaining after this commit for our 
> out-of-tree target. More over, address sanitizer is complaining about an 
> invalid access.
> 
> ==50770==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000008 (pc 0x000102e7e42e bp 0x700002f37170 sp 0x700002f36f60 T9)
> ==50770==The signal is caused by a READ memory access.
> ==50770==Hint: address points to the zero page.
>      #0 0x102e7e42d in (anonymous namespace)::MachineCopyPropagation::forwardUses(llvm::MachineInstr&) MachineCopyPropagation.cpp:596
>      #1 0x102e79c22 in (anonymous namespace)::MachineCopyPropagation::runOnMachineFunction(llvm::MachineFunction&) MachineCopyPropagation.cpp:806
> 
> 
> The MachineVerifier error is:
> 
> *** Bad machine code: Instruction ending live segment doesn't read the register ***
> 
> 
> Could you revert while we investigate?
> 
> Thanks,
> -Quentin
>> On Oct 2, 2017, at 3:01 PM, Geoff Berry via llvm-commits 
>> <llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>> wrote:
>>
>> Author: gberry
>> Date: Mon Oct  2 15:01:37 2017
>> New Revision: 314729
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=314729&view=rev
>> Log:
>> Re-enable "[MachineCopyPropagation] Extend pass to do COPY source 
>> forwarding"
>>
>> Issues addressed since original review:
>> - Avoid bug in regalloc greedy/machine verifier when forwarding to use
>>  in an instruction that re-defines the same virtual register.
>> - Fixed bug when forwarding to use in EarlyClobber instruction slot.
>> - Fixed incorrect forwarding to register definitions that showed up in
>>  explicit_uses() iterator (e.g. in INLINEASM).
>> - Moved removal of dead instructions found by
>>  LiveIntervals::shrinkToUses() outside of loop iterating over
>>  instructions to avoid instructions being deleted while pointed to by
>>  iterator.
>> - Fixed ARMLoadStoreOptimizer bug exposed by this change in r311907.
>> - The pass no longer forwards COPYs to physical register uses, since
>>  doing so can break code that implicitly relies on the physical
>>  register number of the use.
>> - The pass no longer forwards COPYs to undef uses, since doing so
>>  can break the machine verifier by creating LiveRanges that don't
>>  end on a use (since the undef operand is not considered a use).
>>
>>  [MachineCopyPropagation] Extend pass to do COPY source forwarding
>>
>>  This change extends MachineCopyPropagation to do COPY source forwarding.
>>
>>  This change also extends the MachineCopyPropagation pass to be able to
>>  be run during register allocation, after physical registers have been
>>  assigned, but before the virtual registers have been re-written, which
>>  allows it to remove virtual register COPY LiveIntervals that become dead
>>  through the forwarding of all of their uses.
>>
>> Modified:
>>    llvm/trunk/include/llvm/CodeGen/Passes.h
>>    llvm/trunk/include/llvm/InitializePasses.h
>>    llvm/trunk/lib/CodeGen/CodeGen.cpp
>>    llvm/trunk/lib/CodeGen/MachineCopyPropagation.cpp
>>    llvm/trunk/lib/CodeGen/TargetPassConfig.cpp
>>    llvm/trunk/test/CodeGen/AArch64/aarch64-fold-lslfast.ll
>>    llvm/trunk/test/CodeGen/AArch64/arm64-AdvSIMD-Scalar.ll
>>    llvm/trunk/test/CodeGen/AArch64/arm64-zero-cycle-regmov.ll
>>    llvm/trunk/test/CodeGen/AArch64/f16-instructions.ll
>>    llvm/trunk/test/CodeGen/AArch64/flags-multiuse.ll
>>    llvm/trunk/test/CodeGen/AArch64/merge-store-dependency.ll
>>    llvm/trunk/test/CodeGen/AArch64/neg-imm.ll
>>    llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll
>>    llvm/trunk/test/CodeGen/AMDGPU/mad-mix.ll
>>    llvm/trunk/test/CodeGen/AMDGPU/ret.ll
>>    llvm/trunk/test/CodeGen/ARM/atomic-op.ll
>>    llvm/trunk/test/CodeGen/ARM/intrinsics-overflow.ll
>>    llvm/trunk/test/CodeGen/ARM/swifterror.ll
>>    llvm/trunk/test/CodeGen/Mips/llvm-ir/sub.ll
>>    llvm/trunk/test/CodeGen/PowerPC/fma-mutate.ll
>>    llvm/trunk/test/CodeGen/PowerPC/gpr-vsr-spill.ll
>>    llvm/trunk/test/CodeGen/PowerPC/inlineasm-i64-reg.ll
>>    llvm/trunk/test/CodeGen/PowerPC/opt-li-add-to-addi.ll
>>    llvm/trunk/test/CodeGen/PowerPC/tail-dup-layout.ll
>>    llvm/trunk/test/CodeGen/SPARC/atomics.ll
>>    llvm/trunk/test/CodeGen/Thumb/thumb-shrink-wrapping.ll
>>    llvm/trunk/test/CodeGen/X86/2006-03-01-InstrSchedBug.ll
>>    llvm/trunk/test/CodeGen/X86/arg-copy-elide.ll
>>    llvm/trunk/test/CodeGen/X86/avg.ll
>>    llvm/trunk/test/CodeGen/X86/avx-load-store.ll
>>    llvm/trunk/test/CodeGen/X86/avx512-bugfix-25270.ll
>>    llvm/trunk/test/CodeGen/X86/avx512-calling-conv.ll
>>    llvm/trunk/test/CodeGen/X86/avx512-mask-op.ll
>>    llvm/trunk/test/CodeGen/X86/avx512-schedule.ll
>>    llvm/trunk/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
>>    llvm/trunk/test/CodeGen/X86/buildvec-insertvec.ll
>>    llvm/trunk/test/CodeGen/X86/combine-fcopysign.ll
>>    llvm/trunk/test/CodeGen/X86/combine-shl.ll
>>    llvm/trunk/test/CodeGen/X86/complex-fastmath.ll
>>    llvm/trunk/test/CodeGen/X86/divide-by-constant.ll
>>    llvm/trunk/test/CodeGen/X86/fmaxnum.ll
>>    llvm/trunk/test/CodeGen/X86/fmf-flags.ll
>>    llvm/trunk/test/CodeGen/X86/fminnum.ll
>>    llvm/trunk/test/CodeGen/X86/fp128-i128.ll
>>    llvm/trunk/test/CodeGen/X86/haddsub-2.ll
>>    llvm/trunk/test/CodeGen/X86/haddsub-undef.ll
>>    llvm/trunk/test/CodeGen/X86/half.ll
>>    llvm/trunk/test/CodeGen/X86/inline-asm-fpstack.ll
>>    llvm/trunk/test/CodeGen/X86/ipra-local-linkage.ll
>>    llvm/trunk/test/CodeGen/X86/localescape.ll
>>    llvm/trunk/test/CodeGen/X86/machine-cp.ll
>>    llvm/trunk/test/CodeGen/X86/mul-i1024.ll
>>    llvm/trunk/test/CodeGen/X86/mul-i512.ll
>>    llvm/trunk/test/CodeGen/X86/mul128.ll
>>    llvm/trunk/test/CodeGen/X86/mulvi32.ll
>>    llvm/trunk/test/CodeGen/X86/pmul.ll
>>    llvm/trunk/test/CodeGen/X86/powi.ll
>>    llvm/trunk/test/CodeGen/X86/pr11334.ll
>>    llvm/trunk/test/CodeGen/X86/pr29112.ll
>>    llvm/trunk/test/CodeGen/X86/psubus.ll
>>    llvm/trunk/test/CodeGen/X86/shrink-wrap-chkstk.ll
>>    llvm/trunk/test/CodeGen/X86/sqrt-fastmath.ll
>>    llvm/trunk/test/CodeGen/X86/sse1.ll
>>    llvm/trunk/test/CodeGen/X86/sse3-avx-addsub-2.ll
>>    llvm/trunk/test/CodeGen/X86/statepoint-live-in.ll
>>    llvm/trunk/test/CodeGen/X86/statepoint-stack-usage.ll
>>    llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll
>>    llvm/trunk/test/CodeGen/X86/vec_int_to_fp.ll
>>    llvm/trunk/test/CodeGen/X86/vec_minmax_sint.ll
>>    llvm/trunk/test/CodeGen/X86/vec_shift4.ll
>>    llvm/trunk/test/CodeGen/X86/vector-blend.ll
>>    llvm/trunk/test/CodeGen/X86/vector-idiv-sdiv-128.ll
>>    llvm/trunk/test/CodeGen/X86/vector-idiv-udiv-128.ll
>>    llvm/trunk/test/CodeGen/X86/vector-mul.ll
>>    llvm/trunk/test/CodeGen/X86/vector-rotate-128.ll
>>    llvm/trunk/test/CodeGen/X86/vector-sext.ll
>>    llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll
>>    llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll
>>    llvm/trunk/test/CodeGen/X86/vector-shift-shl-128.ll
>>    llvm/trunk/test/CodeGen/X86/vector-shuffle-combining.ll
>>    llvm/trunk/test/CodeGen/X86/vector-trunc-math.ll
>>    llvm/trunk/test/CodeGen/X86/vector-zext.ll
>>    llvm/trunk/test/CodeGen/X86/vselect-minmax.ll
>>    llvm/trunk/test/CodeGen/X86/widen_conv-3.ll
>>    llvm/trunk/test/CodeGen/X86/widen_conv-4.ll
>>    llvm/trunk/test/CodeGen/X86/x86-interleaved-access.ll
>>    llvm/trunk/test/CodeGen/X86/x86-shrink-wrap-unwind.ll
>>    llvm/trunk/test/CodeGen/X86/x86-shrink-wrapping.ll
>>    llvm/trunk/test/DebugInfo/X86/live-debug-variables.ll
>>    llvm/trunk/test/DebugInfo/X86/spill-nospill.ll
>>
>> Modified: llvm/trunk/include/llvm/CodeGen/Passes.h
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/Passes.h?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/include/llvm/CodeGen/Passes.h (original)
>> +++ llvm/trunk/include/llvm/CodeGen/Passes.h Mon Oct  2 15:01:37 2017
>> @@ -278,6 +278,11 @@ namespace llvm {
>>   /// MachineSinking - This pass performs sinking on machine instructions.
>>   extern char &MachineSinkingID;
>>
>> +  /// MachineCopyPropagationPreRegRewrite - This pass performs copy 
>> propagation
>> +  /// on machine instructions after register allocation but before 
>> virtual
>> +  /// register re-writing..
>> +  extern char &MachineCopyPropagationPreRegRewriteID;
>> +
>>   /// MachineCopyPropagation - This pass performs copy propagation on
>>   /// machine instructions.
>>   extern char &MachineCopyPropagationID;
>>
>> Modified: llvm/trunk/include/llvm/InitializePasses.h
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/InitializePasses.h?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/include/llvm/InitializePasses.h (original)
>> +++ llvm/trunk/include/llvm/InitializePasses.h Mon Oct  2 15:01:37 2017
>> @@ -233,6 +233,7 @@ void initializeMachineBranchProbabilityI
>> void initializeMachineCSEPass(PassRegistry&);
>> void initializeMachineCombinerPass(PassRegistry&);
>> void initializeMachineCopyPropagationPass(PassRegistry&);
>> +void initializeMachineCopyPropagationPreRegRewritePass(PassRegistry&);
>> void initializeMachineDominanceFrontierPass(PassRegistry&);
>> void initializeMachineDominatorTreePass(PassRegistry&);
>> void initializeMachineFunctionPrinterPassPass(PassRegistry&);
>>
>> Modified: llvm/trunk/lib/CodeGen/CodeGen.cpp
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/CodeGen.cpp?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/CodeGen/CodeGen.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/CodeGen.cpp Mon Oct  2 15:01:37 2017
>> @@ -53,6 +53,7 @@ void llvm::initializeCodeGen(PassRegistr
>>   initializeMachineCSEPass(Registry);
>>   initializeMachineCombinerPass(Registry);
>>   initializeMachineCopyPropagationPass(Registry);
>> +  initializeMachineCopyPropagationPreRegRewritePass(Registry);
>>   initializeMachineDominatorTreePass(Registry);
>>   initializeMachineFunctionPrinterPassPass(Registry);
>>   initializeMachineLICMPass(Registry);
>>
>> Modified: llvm/trunk/lib/CodeGen/MachineCopyPropagation.cpp
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCopyPropagation.cpp?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/CodeGen/MachineCopyPropagation.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/MachineCopyPropagation.cpp Mon Oct  2 
>> 15:01:37 2017
>> @@ -7,25 +7,71 @@
>> //
>> //===----------------------------------------------------------------------===//
>> //
>> -// This is an extremely simple MachineInstr-level copy propagation pass.
>> +// This is a simple MachineInstr-level copy forwarding pass.  It may 
>> be run at
>> +// two places in the codegen pipeline:
>> +//   - After register allocation but before virtual registers have 
>> been remapped
>> +//     to physical registers.
>> +//   - After physical register remapping.
>> +//
>> +// The optimizations done vary slightly based on whether virtual 
>> registers are
>> +// still present.  In both cases, this pass forwards the source of 
>> COPYs to the
>> +// users of their destinations when doing so is legal.  For example:
>> +//
>> +//   %vreg1 = COPY %vreg0
>> +//   ...
>> +//   ... = OP %vreg1
>> +//
>> +// If
>> +//   - the physical register assigned to %vreg0 has not been 
>> clobbered by the
>> +//     time of the use of %vreg1
>> +//   - the register class constraints are satisfied
>> +//   - the COPY def is the only value that reaches OP
>> +// then this pass replaces the above with:
>> +//
>> +//   %vreg1 = COPY %vreg0
>> +//   ...
>> +//   ... = OP %vreg0
>> +//
>> +// and updates the relevant state required by VirtRegMap (e.g. 
>> LiveIntervals).
>> +// COPYs whose LiveIntervals become dead as a result of this 
>> forwarding (i.e. if
>> +// all uses of %vreg1 are changed to %vreg0) are removed.
>> +//
>> +// When being run with only physical registers, this pass will also 
>> remove some
>> +// redundant COPYs.  For example:
>> +//
>> +//    %R1 = COPY %R0
>> +//    ... // No clobber of %R1
>> +//    %R0 = COPY %R1 <<< Removed
>> +//
>> +// or
>> +//
>> +//    %R1 = COPY %R0
>> +//    ... // No clobber of %R0
>> +//    %R1 = COPY %R0 <<< Removed
>> //
>> //===----------------------------------------------------------------------===//
>>
>> +#include "LiveDebugVariables.h"
>> #include "llvm/ADT/DenseMap.h"
>> #include "llvm/ADT/STLExtras.h"
>> #include "llvm/ADT/SetVector.h"
>> #include "llvm/ADT/SmallVector.h"
>> #include "llvm/ADT/Statistic.h"
>> #include "llvm/ADT/iterator_range.h"
>> +#include "llvm/CodeGen/LiveRangeEdit.h"
>> +#include "llvm/CodeGen/LiveStackAnalysis.h"
>> #include "llvm/CodeGen/MachineBasicBlock.h"
>> #include "llvm/CodeGen/MachineFunction.h"
>> #include "llvm/CodeGen/MachineFunctionPass.h"
>> #include "llvm/CodeGen/MachineInstr.h"
>> #include "llvm/CodeGen/MachineOperand.h"
>> #include "llvm/CodeGen/MachineRegisterInfo.h"
>> +#include "llvm/CodeGen/Passes.h"
>> +#include "llvm/CodeGen/VirtRegMap.h"
>> #include "llvm/MC/MCRegisterInfo.h"
>> #include "llvm/Pass.h"
>> #include "llvm/Support/Debug.h"
>> +#include "llvm/Support/DebugCounter.h"
>> #include "llvm/Support/raw_ostream.h"
>> #include "llvm/Target/TargetInstrInfo.h"
>> #include "llvm/Target/TargetRegisterInfo.h"
>> @@ -38,6 +84,9 @@ using namespace llvm;
>> #define DEBUG_TYPE "machine-cp"
>>
>> STATISTIC(NumDeletes, "Number of dead copies deleted");
>> +STATISTIC(NumCopyForwards, "Number of copy uses forwarded");
>> +DEBUG_COUNTER(FwdCounter, "machine-cp-fwd",
>> +              "Controls which register COPYs are forwarded");
>>
>> namespace {
>>
>> @@ -45,19 +94,42 @@ using RegList = SmallVector<unsigned, 4>
>> using SourceMap = DenseMap<unsigned, RegList>;
>> using Reg2MIMap = DenseMap<unsigned, MachineInstr *>;
>>
>> -  class MachineCopyPropagation : public MachineFunctionPass {
>> +  class MachineCopyPropagation : public MachineFunctionPass,
>> +                                 private LiveRangeEdit::Delegate {
>>     const TargetRegisterInfo *TRI;
>>     const TargetInstrInfo *TII;
>> -    const MachineRegisterInfo *MRI;
>> +    MachineRegisterInfo *MRI;
>> +    MachineFunction *MF;
>> +    SlotIndexes *Indexes;
>> +    LiveIntervals *LIS;
>> +    const VirtRegMap *VRM;
>> +    // True if this pass being run before virtual registers are 
>> remapped to
>> +    // physical ones.
>> +    bool PreRegRewrite;
>> +    bool NoSubRegLiveness;
>> +
>> +  protected:
>> +    MachineCopyPropagation(char &ID, bool PreRegRewrite)
>> +        : MachineFunctionPass(ID), PreRegRewrite(PreRegRewrite) {}
>>
>>   public:
>>     static char ID; // Pass identification, replacement for typeid
>>
>> -    MachineCopyPropagation() : MachineFunctionPass(ID) {
>> +    MachineCopyPropagation() : MachineCopyPropagation(ID, false) {
>>       initializeMachineCopyPropagationPass(*PassRegistry::getPassRegistry());
>>     }
>>
>>     void getAnalysisUsage(AnalysisUsage &AU) const override {
>> +      if (PreRegRewrite) {
>> +        AU.addRequired<SlotIndexes>();
>> +        AU.addPreserved<SlotIndexes>();
>> +        AU.addRequired<LiveIntervals>();
>> +        AU.addPreserved<LiveIntervals>();
>> +        AU.addRequired<VirtRegMap>();
>> +        AU.addPreserved<VirtRegMap>();
>> +        AU.addPreserved<LiveDebugVariables>();
>> +        AU.addPreserved<LiveStacks>();
>> +      }
>>       AU.setPreservesCFG();
>>       MachineFunctionPass::getAnalysisUsage(AU);
>>     }
>> @@ -65,6 +137,10 @@ using Reg2MIMap = DenseMap<unsigned, Mac
>>     bool runOnMachineFunction(MachineFunction &MF) override;
>>
>>     MachineFunctionProperties getRequiredProperties() const override {
>> +      if (PreRegRewrite)
>> +        return MachineFunctionProperties()
>> +            .set(MachineFunctionProperties::Property::NoPHIs)
>> +            .set(MachineFunctionProperties::Property::TracksLiveness);
>>       return MachineFunctionProperties().set(
>>           MachineFunctionProperties::Property::NoVRegs);
>>     }
>> @@ -74,9 +150,33 @@ using Reg2MIMap = DenseMap<unsigned, Mac
>>     void ReadRegister(unsigned Reg);
>>     void CopyPropagateBlock(MachineBasicBlock &MBB);
>>     bool eraseIfRedundant(MachineInstr &Copy, unsigned Src, unsigned Def);
>> +    unsigned getPhysReg(unsigned Reg, unsigned SubReg);
>> +    unsigned getPhysReg(const MachineOperand &Opnd) {
>> +      return getPhysReg(Opnd.getReg(), Opnd.getSubReg());
>> +    }
>> +    unsigned getFullPhysReg(const MachineOperand &Opnd) {
>> +      return getPhysReg(Opnd.getReg(), 0);
>> +    }
>> +    void forwardUses(MachineInstr &MI);
>> +    bool isForwardableRegClassCopy(const MachineInstr &Copy,
>> +                                   const MachineInstr &UseI);
>> +    std::tuple<unsigned, unsigned, bool>
>> +    checkUseSubReg(const MachineOperand &CopySrc, const 
>> MachineOperand &MOUse);
>> +    bool hasImplicitOverlap(const MachineInstr &MI, const 
>> MachineOperand &Use);
>> +    void narrowRegClass(const MachineInstr &MI, const MachineOperand 
>> &MOUse,
>> +                        unsigned NewUseReg, unsigned NewUseSubReg);
>> +    void updateForwardedCopyLiveInterval(const MachineInstr &Copy,
>> +                                         const MachineInstr &UseMI,
>> +                                         bool UseIsEarlyClobber,
>> +                                         unsigned OrigUseReg,
>> +                                         unsigned NewUseReg,
>> +                                         unsigned NewUseSubReg);
>> +    /// LiveRangeEdit callback for eliminateDeadDefs().
>> +    void LRE_WillEraseInstruction(MachineInstr *MI) override;
>>
>>     /// Candidates for deletion.
>>     SmallSetVector<MachineInstr*, 8> MaybeDeadCopies;
>> +    SmallVector<MachineInstr*, 8> ShrunkDeadInsts;
>>
>>     /// Def -> available copies map.
>>     Reg2MIMap AvailCopyMap;
>> @@ -90,6 +190,14 @@ using Reg2MIMap = DenseMap<unsigned, Mac
>>     bool Changed;
>>   };
>>
>> +  class MachineCopyPropagationPreRegRewrite : public 
>> MachineCopyPropagation {
>> +  public:
>> +    static char ID; // Pass identification, replacement for typeid
>> +    MachineCopyPropagationPreRegRewrite()
>> +        : MachineCopyPropagation(ID, true) {
>> + 
>>      initializeMachineCopyPropagationPreRegRewritePass(*PassRegistry::getPassRegistry());
>> +    }
>> +  };
>> } // end anonymous namespace
>>
>> char MachineCopyPropagation::ID = 0;
>> @@ -99,6 +207,29 @@ char &llvm::MachineCopyPropagationID = M
>> INITIALIZE_PASS(MachineCopyPropagation, DEBUG_TYPE,
>>                 "Machine Copy Propagation Pass", false, false)
>>
>> +/// We have two separate passes that are very similar, the only 
>> difference being
>> +/// where they are meant to be run in the pipeline.  This is done for 
>> several
>> +/// reasons:
>> +/// - the two passes have different dependencies
>> +/// - some targets want to disable the later run of this pass, but 
>> not the
>> +///   earlier one (e.g. NVPTX and WebAssembly)
>> +/// - it allows for easier debugging via llc
>> +
>> +char MachineCopyPropagationPreRegRewrite::ID = 0;
>> +char &llvm::MachineCopyPropagationPreRegRewriteID = 
>> MachineCopyPropagationPreRegRewrite::ID;
>> +
>> +INITIALIZE_PASS_BEGIN(MachineCopyPropagationPreRegRewrite,
>> +                      "machine-cp-prerewrite",
>> +                      "Machine Copy Propagation Pre-Register Rewrite 
>> Pass",
>> +                      false, false)
>> +INITIALIZE_PASS_DEPENDENCY(SlotIndexes)
>> +INITIALIZE_PASS_DEPENDENCY(LiveIntervals)
>> +INITIALIZE_PASS_DEPENDENCY(VirtRegMap)
>> +INITIALIZE_PASS_END(MachineCopyPropagationPreRegRewrite,
>> +                    "machine-cp-prerewrite",
>> +                    "Machine Copy Propagation Pre-Register Rewrite 
>> Pass", false,
>> +                    false)
>> +
>> /// Remove any entry in \p Map where the register is a subregister or 
>> equal to
>> /// a register contained in \p Regs.
>> static void removeRegsFromMap(Reg2MIMap &Map, const RegList &Regs,
>> @@ -139,6 +270,10 @@ void MachineCopyPropagation::ClobberRegi
>> }
>>
>> void MachineCopyPropagation::ReadRegister(unsigned Reg) {
>> +  // We don't track MaybeDeadCopies when running pre-VirtRegRewriter.
>> +  if (PreRegRewrite)
>> +    return;
>> +
>>   // If 'Reg' is defined by a copy, the copy is no longer a candidate
>>   // for elimination.
>>   for (MCRegAliasIterator AI(Reg, TRI, true); AI.isValid(); ++AI) {
>> @@ -170,6 +305,46 @@ static bool isNopCopy(const MachineInstr
>>   return SubIdx == TRI->getSubRegIndex(PreviousDef, Def);
>> }
>>
>> +/// Return the physical register assigned to \p Reg if it is a 
>> virtual register,
>> +/// otherwise just return the physical reg from the operand itself.
>> +///
>> +/// If \p SubReg is 0 then return the full physical register assigned 
>> to the
>> +/// virtual register ignoring subregs.  If we aren't tracking sub-reg 
>> liveness
>> +/// then we need to use this to be more conservative with clobbers by 
>> killing
>> +/// all super reg and their sub reg COPYs as well.  This is to 
>> prevent COPY
>> +/// forwarding in cases like the following:
>> +///
>> +///    %vreg2 = COPY %vreg1:sub1
>> +///    %vreg3 = COPY %vreg1:sub0
>> +///    ...    = OP1 %vreg2
>> +///    ...    = OP2 %vreg3
>> +///
>> +/// After forward %vreg2 (assuming this is the last use of %vreg1) and
>> +/// VirtRegRewriter adding kill markers we have:
>> +///
>> +///    %vreg3 = COPY %vreg1:sub0
>> +///    ...    = OP1 %vreg1:sub1<kill>
>> +///    ...    = OP2 %vreg3
>> +///
>> +/// If %vreg3 is assigned to a sub-reg of %vreg1, then after 
>> rewriting we have:
>> +///
>> +///    ...     = OP1 R0:sub1, R0<imp-use,kill>
>> +///    ...     = OP2 R0:sub0
>> +///
>> +/// and the use of R0 by OP2 will not have a valid definition.
>> +unsigned MachineCopyPropagation::getPhysReg(unsigned Reg, unsigned 
>> SubReg) {
>> +
>> +  // Physical registers cannot have subregs.
>> +  if (!TargetRegisterInfo::isVirtualRegister(Reg))
>> +    return Reg;
>> +
>> +  assert(PreRegRewrite && "Unexpected virtual register encountered");
>> +  Reg = VRM->getPhys(Reg);
>> +  if (SubReg && !NoSubRegLiveness)
>> +    Reg = TRI->getSubReg(Reg, SubReg);
>> +  return Reg;
>> +}
>> +
>> /// Remove instruction \p Copy if there exists a previous copy that 
>> copies the
>> /// register \p Src to the register \p Def; This may happen indirectly by
>> /// copying the super registers.
>> @@ -207,6 +382,397 @@ bool MachineCopyPropagation::eraseIfRedu
>>   return true;
>> }
>>
>> +
>> +/// Decide whether we should forward the destination of \param Copy 
>> to its use
>> +/// in \param UseI based on the register class of the Copy operands. 
>>  Same-class
>> +/// COPYs are always accepted by this function, but cross-class COPYs 
>> are only
>> +/// accepted if they are forwarded to another COPY with the operand 
>> register
>> +/// classes reversed.  For example:
>> +///
>> +///   RegClassA = COPY RegClassB  // Copy parameter
>> +///   ...
>> +///   RegClassB = COPY RegClassA  // UseI parameter
>> +///
>> +/// which after forwarding becomes
>> +///
>> +///   RegClassA = COPY RegClassB
>> +///   ...
>> +///   RegClassB = COPY RegClassB
>> +///
>> +/// so we have reduced the number of cross-class COPYs and potentially
>> +/// introduced a no COPY that can be removed.
>> +bool MachineCopyPropagation::isForwardableRegClassCopy(
>> +    const MachineInstr &Copy, const MachineInstr &UseI) {
>> +  auto isCross = [&](const MachineOperand &Dst, const MachineOperand 
>> &Src) {
>> +    unsigned DstReg = Dst.getReg();
>> +    unsigned SrcPhysReg = getPhysReg(Src);
>> +    const TargetRegisterClass *DstRC;
>> +    if (TargetRegisterInfo::isVirtualRegister(DstReg)) {
>> +      DstRC = MRI->getRegClass(DstReg);
>> +      unsigned DstSubReg = Dst.getSubReg();
>> +      if (DstSubReg)
>> +        SrcPhysReg = TRI->getMatchingSuperReg(SrcPhysReg, DstSubReg, 
>> DstRC);
>> +    } else
>> +      DstRC = TRI->getMinimalPhysRegClass(DstReg);
>> +
>> +    return !DstRC->contains(SrcPhysReg);
>> +  };
>> +
>> +  const MachineOperand &CopyDst = Copy.getOperand(0);
>> +  const MachineOperand &CopySrc = Copy.getOperand(1);
>> +
>> +  if (!isCross(CopyDst, CopySrc))
>> +    return true;
>> +
>> +  if (!UseI.isCopy())
>> +    return false;
>> +
>> +  assert(getFullPhysReg(UseI.getOperand(1)) == getFullPhysReg(CopyDst));
>> +  return !isCross(UseI.getOperand(0), CopySrc);
>> +}
>> +
>> +/// Check that the subregs on the copy source operand (\p CopySrc) 
>> and the use
>> +/// operand to be forwarded to (\p MOUse) are compatible with doing the
>> +/// forwarding.  Also computes the new register and subregister to be 
>> used in
>> +/// the forwarded-to instruction.
>> +std::tuple<unsigned, unsigned, bool> 
>> MachineCopyPropagation::checkUseSubReg(
>> +    const MachineOperand &CopySrc, const MachineOperand &MOUse) {
>> +  unsigned NewUseReg = CopySrc.getReg();
>> +  unsigned NewUseSubReg;
>> +
>> +  if (TargetRegisterInfo::isPhysicalRegister(NewUseReg)) {
>> +    // If MOUse is a virtual reg, we need to apply it to the new 
>> physical reg
>> +    // we're going to replace it with.
>> +    if (MOUse.getSubReg())
>> +      NewUseReg = TRI->getSubReg(NewUseReg, MOUse.getSubReg());
>> +    // If the original use subreg isn't valid on the new src reg, we 
>> can't
>> +    // forward it here.
>> +    if (!NewUseReg)
>> +      return std::make_tuple(0, 0, false);
>> +    NewUseSubReg = 0;
>> +  } else {
>> +    // %v1 = COPY %v2:sub1
>> +    //    USE %v1:sub2
>> +    // The new use is %v2:sub1:sub2
>> +    NewUseSubReg =
>> +        TRI->composeSubRegIndices(CopySrc.getSubReg(), 
>> MOUse.getSubReg());
>> +    // Check that NewUseSubReg is valid on NewUseReg
>> +    if (NewUseSubReg &&
>> +        !TRI->getSubClassWithSubReg(MRI->getRegClass(NewUseReg), 
>> NewUseSubReg))
>> +      return std::make_tuple(0, 0, false);
>> +  }
>> +
>> +  return std::make_tuple(NewUseReg, NewUseSubReg, true);
>> +}
>> +
>> +/// Check that \p MI does not have implicit uses that overlap with 
>> it's \p Use
>> +/// operand (the register being replaced), since these can sometimes be
>> +/// implicitly tied to other operands.  For example, on AMDGPU:
>> +///
>> +/// V_MOVRELS_B32_e32 %VGPR2, %M0<imp-use>, %EXEC<imp-use>, 
>> %VGPR2_VGPR3_VGPR4_VGPR5<imp-use>
>> +///
>> +/// the %VGPR2 is implicitly tied to the larger reg operand, but we 
>> have no
>> +/// way of knowing we need to update the latter when updating the former.
>> +bool MachineCopyPropagation::hasImplicitOverlap(const MachineInstr &MI,
>> +                                                const MachineOperand 
>> &Use) {
>> +  if (!TargetRegisterInfo::isPhysicalRegister(Use.getReg()))
>> +    return false;
>> +
>> +  for (const MachineOperand &MIUse : MI.uses())
>> +    if (&MIUse != &Use && MIUse.isReg() && MIUse.isImplicit() &&
>> +        TRI->regsOverlap(Use.getReg(), MIUse.getReg()))
>> +      return true;
>> +
>> +  return false;
>> +}
>> +
>> +/// Narrow the register class of the forwarded vreg so it matches any
>> +/// instruction constraints.  \p MI is the instruction being 
>> forwarded to. \p
>> +/// MOUse is the operand being replaced in \p MI (which hasn't yet 
>> been updated
>> +/// at the time this function is called).  \p NewUseReg and \p 
>> NewUseSubReg are
>> +/// what the \p MOUse will be changed to after forwarding.
>> +///
>> +/// If we are forwarding
>> +///    A:RCA = COPY B:RCB
>> +/// into
>> +///    ... = OP A:RCA
>> +///
>> +/// then we need to narrow the register class of B so that it is a 
>> subclass
>> +/// of RCA so that it meets the instruction register class constraints.
>> +void MachineCopyPropagation::narrowRegClass(const MachineInstr &MI,
>> +                                            const MachineOperand &MOUse,
>> +                                            unsigned NewUseReg,
>> +                                            unsigned NewUseSubReg) {
>> +  if (!TargetRegisterInfo::isVirtualRegister(NewUseReg))
>> +    return;
>> +
>> +  // Make sure the virtual reg class allows the subreg.
>> +  if (NewUseSubReg) {
>> +    const TargetRegisterClass *CurUseRC = MRI->getRegClass(NewUseReg);
>> +    const TargetRegisterClass *NewUseRC =
>> +        TRI->getSubClassWithSubReg(CurUseRC, NewUseSubReg);
>> +    if (CurUseRC != NewUseRC) {
>> +      DEBUG(dbgs() << "MCP: Setting regclass of " << 
>> PrintReg(NewUseReg, TRI)
>> +                   << " to " << TRI->getRegClassName(NewUseRC) << "\n");
>> +      MRI->setRegClass(NewUseReg, NewUseRC);
>> +    }
>> +  }
>> +
>> +  unsigned MOUseOpNo = &MOUse - &MI.getOperand(0);
>> +  const TargetRegisterClass *InstRC =
>> +      TII->getRegClass(MI.getDesc(), MOUseOpNo, TRI, *MF);
>> +  if (InstRC) {
>> +    const TargetRegisterClass *CurUseRC = MRI->getRegClass(NewUseReg);
>> +    if (NewUseSubReg)
>> +      InstRC = TRI->getMatchingSuperRegClass(CurUseRC, InstRC, 
>> NewUseSubReg);
>> +    if (!InstRC->hasSubClassEq(CurUseRC)) {
>> +      const TargetRegisterClass *NewUseRC =
>> +          TRI->getCommonSubClass(InstRC, CurUseRC);
>> +      DEBUG(dbgs() << "MCP: Setting regclass of " << 
>> PrintReg(NewUseReg, TRI)
>> +                   << " to " << TRI->getRegClassName(NewUseRC) << "\n");
>> +      MRI->setRegClass(NewUseReg, NewUseRC);
>> +    }
>> +  }
>> +}
>> +
>> +/// Update the LiveInterval information to reflect the destination of 
>> \p Copy
>> +/// being forwarded to a use in \p UseMI.  \p OrigUseReg is the 
>> register being
>> +/// forwarded through. It should be the destination register of \p 
>> Copy and has
>> +/// already been replaced in \p UseMI at the point this function is 
>> called.  \p
>> +/// NewUseReg and \p NewUseSubReg are the register and subregister being
>> +/// forwarded.  They should be the source register of the \p Copy and 
>> should be
>> +/// the value of the \p UseMI operand being forwarded at the point 
>> this function
>> +/// is called.  \p UseIsEarlyClobber is true if the use being 
>> re-written is in
>> +/// the EarlyClobber slot index (as opposed to the register slot of most
>> +/// register operands).
>> +void MachineCopyPropagation::updateForwardedCopyLiveInterval(
>> +    const MachineInstr &Copy, const MachineInstr &UseMI, bool 
>> UseIsEarlyClobber,
>> +    unsigned OrigUseReg, unsigned NewUseReg, unsigned NewUseSubReg) {
>> +
>> +  assert(TRI->isSubRegisterEq(getPhysReg(OrigUseReg, 0),
>> +                              getFullPhysReg(Copy.getOperand(0))) &&
>> +         "OrigUseReg mismatch");
>> +  assert(TRI->isSubRegisterEq(getFullPhysReg(Copy.getOperand(1)),
>> +                              getPhysReg(NewUseReg, 0)) &&
>> +         "NewUseReg mismatch");
>> +
>> +  // Extend live range starting from COPY early-clobber slot, since that
>> +  // is where the original src live range ends.
>> +  SlotIndex CopyUseIdx =
>> +      Indexes->getInstructionIndex(Copy).getRegSlot(true 
>> /*=EarlyClobber*/);
>> +  SlotIndex UseEndIdx =
>> +      Indexes->getInstructionIndex(UseMI).getRegSlot(UseIsEarlyClobber);
>> +  if (TargetRegisterInfo::isVirtualRegister(NewUseReg)) {
>> +    LiveInterval &LI = LIS->getInterval(NewUseReg);
>> +    LI.extendInBlock(CopyUseIdx, UseEndIdx);
>> +    LaneBitmask UseMask = TRI->getSubRegIndexLaneMask(NewUseSubReg);
>> +    for (auto &S : LI.subranges())
>> +      if ((S.LaneMask & UseMask).any() && S.find(CopyUseIdx))
>> +        S.extendInBlock(CopyUseIdx, UseEndIdx);
>> +  } else {
>> +    assert(NewUseSubReg == 0 && "Unexpected subreg on physical 
>> register!");
>> +    for (MCRegUnitIterator UI(NewUseReg, TRI); UI.isValid(); ++UI) {
>> +      LiveRange &LR = LIS->getRegUnit(*UI);
>> +      LR.extendInBlock(CopyUseIdx, UseEndIdx);
>> +    }
>> +  }
>> +
>> +  if (!TargetRegisterInfo::isVirtualRegister(OrigUseReg))
>> +    return;
>> +
>> +  // Shrink the live-range for the old use reg if the forwarded use 
>> was it's
>> +  // last use.
>> +  LiveInterval &OrigUseLI = LIS->getInterval(OrigUseReg);
>> +
>> +  // Can happen for undef uses.
>> +  if (OrigUseLI.empty())
>> +    return;
>> +
>> +  SlotIndex CopyDefIdx = Indexes->getInstructionIndex(Copy).getRegSlot();
>> +  const LiveRange::Segment *OrigUseSeg =
>> +      OrigUseLI.getSegmentContaining(CopyDefIdx);
>> +
>> +  // Only shrink if forwarded use is the end of a segment.
>> +  if (OrigUseSeg->end != UseEndIdx)
>> +    return;
>> +
>> +  LIS->shrinkToUses(&OrigUseLI, &ShrunkDeadInsts);
>> +}
>> +
>> +void MachineCopyPropagation::LRE_WillEraseInstruction(MachineInstr *MI) {
>> +  Changed = true;
>> +}
>> +
>> +/// Look for available copies whose destination register is used by 
>> \p MI and
>> +/// replace the use in \p MI with the copy's source register.
>> +void MachineCopyPropagation::forwardUses(MachineInstr &MI) {
>> +  // We can't generally forward uses after virtual registers have 
>> been renamed
>> +  // because some targets generate code that has implicit 
>> dependencies on the
>> +  // physical register numbers.  For example, in PowerPC, when spilling
>> +  // condition code registers, the following code pattern is generated:
>> +  //
>> +  //   %CR7 = COPY %CR0
>> +  //   %R6 = MFOCRF %CR7
>> +  //   %R6 = RLWINM %R6, 29, 31, 31
>> +  //
>> +  // where the shift amount in the RLWINM instruction depends on the 
>> source
>> +  // register number of the MFOCRF instruction.  If we were to 
>> forward %CR0 to
>> +  // the MFOCRF instruction, the shift amount would no longer be correct.
>> +  //
>> +  // FIXME: It may be possible to define a target hook that checks 
>> the register
>> +  // class or user opcode and allows some cases, but prevents cases 
>> like the
>> +  // above from being broken to enable later register copy forwarding.
>> +  if (!PreRegRewrite)
>> +    return;
>> +
>> +  if (AvailCopyMap.empty())
>> +    return;
>> +
>> +  // Look for non-tied explicit vreg uses that have an active COPY
>> +  // instruction that defines the physical register allocated to them.
>> +  // Replace the vreg with the source of the active COPY.
>> +  for (MachineOperand &MOUse : MI.explicit_uses()) {
>> +    // Don't forward into undef use operands since doing so can cause 
>> problems
>> +    // with the machine verifier, since it doesn't treat undef reads 
>> as reads,
>> +    // so we can end up with a live range that ends on an undef read, 
>> leading to
>> +    // an error that the live range doesn't end on a read of the live 
>> range
>> +    // register.
>> +    if (!MOUse.isReg() || MOUse.isTied() || MOUse.isUndef() || 
>> MOUse.isDef())
>> +      continue;
>> +
>> +    unsigned UseReg = MOUse.getReg();
>> +    if (!UseReg)
>> +      continue;
>> +
>> +    // See comment above check for !PreRegRewrite regarding 
>> forwarding changing
>> +    // physical registers.
>> +    if (!TargetRegisterInfo::isVirtualRegister(UseReg))
>> +      continue;
>> +    else {
>> +      // FIXME: Don't forward COPYs to a use that is on an 
>> instruction that
>> +      // re-defines the same virtual register.  This leads to machine
>> +      // verification failures because of a bug in the greedy
>> +      // allocator/verifier.  The bug is that just after greedy 
>> regalloc, we can
>> +      // end up with code that looks like:
>> +      //
>> +      // %vreg1<def> = ...
>> +      // ...
>> +      // ... = %vreg1
>> +      // ...
>> +      // %vreg1<def> = %vreg1
>> +      // ...
>> +      //
>> +      // verifyLiveInterval() accepts this code as valid since it 
>> sees the
>> +      // second def as part of the same live interval component because
>> +      // ConnectedVNInfoEqClasses::Classify() sees this second def as a
>> +      // "two-addr" redefinition, even though the def and source 
>> operands are
>> +      // not tied.
>> +      //
>> +      // If we replace just the second use of %vreg1 in the above 
>> code, then we
>> +      // end up with:
>> +      //
>> +      // ...
>> +      // %vreg1<def> = ...
>> +      // ...
>> +      // ... = %vreg1
>> +      // ...
>> +      // %vreg1<def> = *%vreg2*
>> +      // ...
>> +      //
>> +      // verifyLiveInterval() now rejects this code since it sees 
>> these two def
>> +      // live ranges as being separate components.  To get rid of this
>> +      // restriction on forwarding, regalloc greedy would need to be 
>> fixed to
>> +      // avoid generating code like the first snippet above, as well 
>> as the
>> +      // verifier being fixed to reject such code.
>> +      if (llvm::any_of(MI.defs(), [UseReg](MachineOperand &Def) {
>> +            if (!Def.isReg() || !Def.isDef() || Def.getReg() != UseReg)
>> +              return false;
>> +            // Only tied use/def operands or a subregister def 
>> without the undef
>> +            // flag should result in connected liveranges.
>> +            if (Def.isTied())
>> +              return false;
>> +            if (Def.getSubReg() != 0 && Def.isUndef())
>> +              return false;
>> +            return true;
>> +          }))
>> +        continue;
>> +    }
>> +
>> +    UseReg = VRM->getPhys(UseReg);
>> +
>> +    // Don't forward COPYs via non-allocatable regs since they can have
>> +    // non-standard semantics.
>> +    if (!MRI->isAllocatable(UseReg))
>> +      continue;
>> +
>> +    auto CI = AvailCopyMap.find(UseReg);
>> +    if (CI == AvailCopyMap.end())
>> +      continue;
>> +
>> +    MachineInstr &Copy = *CI->second;
>> +    MachineOperand &CopyDst = Copy.getOperand(0);
>> +    MachineOperand &CopySrc = Copy.getOperand(1);
>> +
>> +    // Don't forward COPYs that are already NOPs due to register 
>> assignment.
>> +    if (getPhysReg(CopyDst) == getPhysReg(CopySrc))
>> +      continue;
>> +
>> +    // FIXME: Don't handle partial uses of wider COPYs yet.
>> +    if (CopyDst.getSubReg() != 0 || UseReg != getPhysReg(CopyDst))
>> +      continue;
>> +
>> +    // Don't forward COPYs of non-allocatable regs unless they are 
>> constant.
>> +    unsigned CopySrcReg = CopySrc.getReg();
>> +    if (TargetRegisterInfo::isPhysicalRegister(CopySrcReg) &&
>> +        !MRI->isAllocatable(CopySrcReg) && 
>> !MRI->isConstantPhysReg(CopySrcReg))
>> +      continue;
>> +
>> +    if (!isForwardableRegClassCopy(Copy, MI))
>> +      continue;
>> +
>> +    unsigned NewUseReg, NewUseSubReg;
>> +    bool SubRegOK;
>> +    std::tie(NewUseReg, NewUseSubReg, SubRegOK) =
>> +        checkUseSubReg(CopySrc, MOUse);
>> +    if (!SubRegOK)
>> +      continue;
>> +
>> +    if (hasImplicitOverlap(MI, MOUse))
>> +      continue;
>> +
>> +    if (!DebugCounter::shouldExecute(FwdCounter))
>> +      continue;
>> +
>> +    DEBUG(dbgs() << "MCP: Replacing "
>> +          << PrintReg(MOUse.getReg(), TRI, MOUse.getSubReg())
>> +          << "\n     with "
>> +          << PrintReg(NewUseReg, TRI, CopySrc.getSubReg())
>> +          << "\n     in "
>> +          << MI
>> +          << "     from "
>> +          << Copy);
>> +
>> +    narrowRegClass(MI, MOUse, NewUseReg, NewUseSubReg);
>> +
>> +    unsigned OrigUseReg = MOUse.getReg();
>> +    MOUse.setReg(NewUseReg);
>> +    MOUse.setSubReg(NewUseSubReg);
>> +
>> +    DEBUG(dbgs() << "MCP: After replacement: " << MI << "\n");
>> +
>> +    if (PreRegRewrite)
>> +      updateForwardedCopyLiveInterval(Copy, MI, MOUse.isEarlyClobber(),
>> +                                      OrigUseReg, NewUseReg, 
>> NewUseSubReg);
>> +    else
>> +      for (MachineInstr &KMI :
>> +             make_range(Copy.getIterator(), std::next(MI.getIterator())))
>> +        KMI.clearRegisterKills(NewUseReg, TRI);
>> +
>> +    ++NumCopyForwards;
>> +    Changed = true;
>> +  }
>> +}
>> +
>> void MachineCopyPropagation::CopyPropagateBlock(MachineBasicBlock &MBB) {
>>   DEBUG(dbgs() << "MCP: CopyPropagateBlock " << MBB.getName() << "\n");
>>
>> @@ -215,12 +781,8 @@ void MachineCopyPropagation::CopyPropaga
>>     ++I;
>>
>>     if (MI->isCopy()) {
>> -      unsigned Def = MI->getOperand(0).getReg();
>> -      unsigned Src = MI->getOperand(1).getReg();
>> -
>> -      assert(!TargetRegisterInfo::isVirtualRegister(Def) &&
>> -             !TargetRegisterInfo::isVirtualRegister(Src) &&
>> -             "MachineCopyPropagation should be run after register 
>> allocation!");
>> +      unsigned Def = getPhysReg(MI->getOperand(0));
>> +      unsigned Src = getPhysReg(MI->getOperand(1));
>>
>>       // The two copies cancel out and the source of the first copy
>>       // hasn't been overridden, eliminate the second one. e.g.
>> @@ -237,8 +799,16 @@ void MachineCopyPropagation::CopyPropaga
>>       //  %ECX<def> = COPY %EAX
>>       // =>
>>       //  %ECX<def> = COPY %EAX
>> -      if (eraseIfRedundant(*MI, Def, Src) || eraseIfRedundant(*MI, 
>> Src, Def))
>> -        continue;
>> +      if (!PreRegRewrite)
>> +        if (eraseIfRedundant(*MI, Def, Src) || eraseIfRedundant(*MI, 
>> Src, Def))
>> +          continue;
>> +
>> +      forwardUses(*MI);
>> +
>> +      // Src may have been changed by forwardUses()
>> +      Src = getPhysReg(MI->getOperand(1));
>> +      unsigned DefClobber = getFullPhysReg(MI->getOperand(0));
>> +      unsigned SrcClobber = getFullPhysReg(MI->getOperand(1));
>>
>>       // If Src is defined by a previous copy, the previous copy cannot be
>>       // eliminated.
>> @@ -255,7 +825,10 @@ void MachineCopyPropagation::CopyPropaga
>>       DEBUG(dbgs() << "MCP: Copy is a deletion candidate: "; MI->dump());
>>
>>       // Copy is now a candidate for deletion.
>> -      if (!MRI->isReserved(Def))
>> +      // Only look for dead COPYs if we're not running just before
>> +      // VirtRegRewriter, since presumably these COPYs will have 
>> already been
>> +      // removed.
>> +      if (!PreRegRewrite && !MRI->isReserved(Def))
>>         MaybeDeadCopies.insert(MI);
>>
>>       // If 'Def' is previously source of another copy, then this 
>> earlier copy's
>> @@ -265,11 +838,11 @@ void MachineCopyPropagation::CopyPropaga
>>       // %xmm2<def> = copy %xmm0
>>       // ...
>>       // %xmm2<def> = copy %xmm9
>> -      ClobberRegister(Def);
>> +      ClobberRegister(DefClobber);
>>       for (const MachineOperand &MO : MI->implicit_operands()) {
>>         if (!MO.isReg() || !MO.isDef())
>>           continue;
>> -        unsigned Reg = MO.getReg();
>> +        unsigned Reg = getFullPhysReg(MO);
>>         if (!Reg)
>>           continue;
>>         ClobberRegister(Reg);
>> @@ -284,13 +857,27 @@ void MachineCopyPropagation::CopyPropaga
>>
>>       // Remember source that's copied to Def. Once it's clobbered, then
>>       // it's no longer available for copy propagation.
>> -      RegList &DestList = SrcMap[Src];
>> -      if (!is_contained(DestList, Def))
>> -        DestList.push_back(Def);
>> +      RegList &DestList = SrcMap[SrcClobber];
>> +      if (!is_contained(DestList, DefClobber))
>> +        DestList.push_back(DefClobber);
>>
>>       continue;
>>     }
>>
>> +    // Clobber any earlyclobber regs first.
>> +    for (const MachineOperand &MO : MI->operands())
>> +      if (MO.isReg() && MO.isEarlyClobber()) {
>> +        unsigned Reg = getFullPhysReg(MO);
>> +        // If we have a tied earlyclobber, that means it is also read 
>> by this
>> +        // instruction, so we need to make sure we don't remove it as 
>> dead
>> +        // later.
>> +        if (MO.isTied())
>> +          ReadRegister(Reg);
>> +        ClobberRegister(Reg);
>> +      }
>> +
>> +    forwardUses(*MI);
>> +
>>     // Not a copy.
>>     SmallVector<unsigned, 2> Defs;
>>     const MachineOperand *RegMask = nullptr;
>> @@ -299,14 +886,11 @@ void MachineCopyPropagation::CopyPropaga
>>         RegMask = &MO;
>>       if (!MO.isReg())
>>         continue;
>> -      unsigned Reg = MO.getReg();
>> +      unsigned Reg = getFullPhysReg(MO);
>>       if (!Reg)
>>         continue;
>>
>> -      assert(!TargetRegisterInfo::isVirtualRegister(Reg) &&
>> -             "MachineCopyPropagation should be run after register 
>> allocation!");
>> -
>> -      if (MO.isDef()) {
>> +      if (MO.isDef() && !MO.isEarlyClobber()) {
>>         Defs.push_back(Reg);
>>         continue;
>>       } else if (MO.readsReg())
>> @@ -358,11 +942,22 @@ void MachineCopyPropagation::CopyPropaga
>>       ClobberRegister(Reg);
>>   }
>>
>> +  // Remove instructions that were made dead by shrinking live 
>> ranges.  Do this
>> +  // after iterating over instructions to avoid instructions changing 
>> while
>> +  // iterating.
>> +  if (!ShrunkDeadInsts.empty()) {
>> +    SmallVector<unsigned, 8> NewRegs;
>> +    LiveRangeEdit(nullptr, NewRegs, *MF, *LIS, nullptr, this)
>> +        .eliminateDeadDefs(ShrunkDeadInsts);
>> +  }
>> +
>>   // If MBB doesn't have successors, delete the copies whose defs are 
>> not used.
>>   // If MBB does have successors, then conservative assume the defs 
>> are live-out
>>   // since we don't want to trust live-in lists.
>>   if (MBB.succ_empty()) {
>>     for (MachineInstr *MaybeDead : MaybeDeadCopies) {
>> +      DEBUG(dbgs() << "MCP: Removing copy due to no live-out succ: ";
>> +            MaybeDead->dump());
>>       assert(!MRI->isReserved(MaybeDead->getOperand(0).getReg()));
>>       MaybeDead->eraseFromParent();
>>       Changed = true;
>> @@ -374,6 +969,7 @@ void MachineCopyPropagation::CopyPropaga
>>   AvailCopyMap.clear();
>>   CopyMap.clear();
>>   SrcMap.clear();
>> +  ShrunkDeadInsts.clear();
>> }
>>
>> bool MachineCopyPropagation::runOnMachineFunction(MachineFunction &MF) {
>> @@ -385,6 +981,13 @@ bool MachineCopyPropagation::runOnMachin
>>   TRI = MF.getSubtarget().getRegisterInfo();
>>   TII = MF.getSubtarget().getInstrInfo();
>>   MRI = &MF.getRegInfo();
>> +  this->MF = &MF;
>> +  if (PreRegRewrite) {
>> +    Indexes = &getAnalysis<SlotIndexes>();
>> +    LIS = &getAnalysis<LiveIntervals>();
>> +    VRM = &getAnalysis<VirtRegMap>();
>> +  }
>> +  NoSubRegLiveness = !MRI->subRegLivenessEnabled();
>>
>>   for (MachineBasicBlock &MBB : MF)
>>     CopyPropagateBlock(MBB);
>>
>> Modified: llvm/trunk/lib/CodeGen/TargetPassConfig.cpp
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/TargetPassConfig.cpp?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/CodeGen/TargetPassConfig.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/TargetPassConfig.cpp Mon Oct  2 15:01:37 2017
>> @@ -88,6 +88,8 @@ static cl::opt<bool> DisableCGP("disable
>>     cl::desc("Disable Codegen Prepare"));
>> static cl::opt<bool> DisableCopyProp("disable-copyprop", cl::Hidden,
>>     cl::desc("Disable Copy Propagation pass"));
>> +static cl::opt<bool> 
>> DisableCopyPropPreRegRewrite("disable-copyprop-prerewrite", cl::Hidden,
>> +    cl::desc("Disable Copy Propagation Pre-Register Re-write pass"));
>> static cl::opt<bool> 
>> DisablePartialLibcallInlining("disable-partial-libcall-inlining",
>>     cl::Hidden, cl::desc("Disable Partial Libcall Inlining"));
>> static cl::opt<bool> EnableImplicitNullChecks(
>> @@ -252,6 +254,9 @@ static IdentifyingPassPtr overridePass(A
>>   if (StandardID == &MachineCopyPropagationID)
>>     return applyDisable(TargetID, DisableCopyProp);
>>
>> +  if (StandardID == &MachineCopyPropagationPreRegRewriteID)
>> +    return applyDisable(TargetID, DisableCopyPropPreRegRewrite);
>> +
>>   return TargetID;
>> }
>>
>> @@ -1064,6 +1069,10 @@ void TargetPassConfig::addOptimizedRegAl
>>     // Allow targets to change the register assignments before rewriting.
>>     addPreRewrite();
>>
>> +    // Copy propagate to forward register uses and try to eliminate 
>> COPYs that
>> +    // were not coalesced.
>> +    addPass(&MachineCopyPropagationPreRegRewriteID);
>> +
>>     // Finally rewrite virtual registers.
>>     addPass(&VirtRegRewriterID);
>>
>>
>> Modified: llvm/trunk/test/CodeGen/AArch64/aarch64-fold-lslfast.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/aarch64-fold-lslfast.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AArch64/aarch64-fold-lslfast.ll (original)
>> +++ llvm/trunk/test/CodeGen/AArch64/aarch64-fold-lslfast.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -9,7 +9,8 @@ define i16 @halfword(%struct.a* %ctx, i3
>> ; CHECK-LABEL: halfword:
>> ; CHECK: ubfx [[REG:x[0-9]+]], x1, #9, #8
>> ; CHECK: ldrh [[REG1:w[0-9]+]], [{{.*}}[[REG2:x[0-9]+]], [[REG]], lsl #1]
>> -; CHECK: strh [[REG1]], [{{.*}}[[REG2]], [[REG]], lsl #1]
>> +; CHECK: mov [[REG3:x[0-9]+]], [[REG2]]
>> +; CHECK: strh [[REG1]], [{{.*}}[[REG3]], [[REG]], lsl #1]
>>   %shr81 = lshr i32 %xor72, 9
>>   %conv82 = zext i32 %shr81 to i64
>>   %idxprom83 = and i64 %conv82, 255
>> @@ -24,7 +25,8 @@ define i32 @word(%struct.b* %ctx, i32 %x
>> ; CHECK-LABEL: word:
>> ; CHECK: ubfx [[REG:x[0-9]+]], x1, #9, #8
>> ; CHECK: ldr [[REG1:w[0-9]+]], [{{.*}}[[REG2:x[0-9]+]], [[REG]], lsl #2]
>> -; CHECK: str [[REG1]], [{{.*}}[[REG2]], [[REG]], lsl #2]
>> +; CHECK: mov [[REG3:x[0-9]+]], [[REG2]]
>> +; CHECK: str [[REG1]], [{{.*}}[[REG3]], [[REG]], lsl #2]
>>   %shr81 = lshr i32 %xor72, 9
>>   %conv82 = zext i32 %shr81 to i64
>>   %idxprom83 = and i64 %conv82, 255
>> @@ -39,7 +41,8 @@ define i64 @doubleword(%struct.c* %ctx,
>> ; CHECK-LABEL: doubleword:
>> ; CHECK: ubfx [[REG:x[0-9]+]], x1, #9, #8
>> ; CHECK: ldr [[REG1:x[0-9]+]], [{{.*}}[[REG2:x[0-9]+]], [[REG]], lsl #3]
>> -; CHECK: str [[REG1]], [{{.*}}[[REG2]], [[REG]], lsl #3]
>> +; CHECK: mov [[REG3:x[0-9]+]], [[REG2]]
>> +; CHECK: str [[REG1]], [{{.*}}[[REG3]], [[REG]], lsl #3]
>>   %shr81 = lshr i32 %xor72, 9
>>   %conv82 = zext i32 %shr81 to i64
>>   %idxprom83 = and i64 %conv82, 255
>>
>> Modified: llvm/trunk/test/CodeGen/AArch64/arm64-AdvSIMD-Scalar.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/arm64-AdvSIMD-Scalar.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AArch64/arm64-AdvSIMD-Scalar.ll (original)
>> +++ llvm/trunk/test/CodeGen/AArch64/arm64-AdvSIMD-Scalar.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -8,15 +8,9 @@ define <2 x i64> @bar(<2 x i64> %a, <2 x
>> ; CHECK: add.2dv[[REG:[0-9]+]], v0, v1
>> ; CHECK: addd[[REG3:[0-9]+]], d[[REG]], d1
>> ; CHECK: subd[[REG2:[0-9]+]], d[[REG]], d1
>> -; Without advanced copy optimization, we end up with cross register
>> -; banks copies that cannot be coalesced.
>> -; CHECK-NOOPT: fmov [[COPY_REG3:x[0-9]+]], d[[REG3]]
>> -; With advanced copy optimization, we end up with just one copy
>> -; to insert the computed high part into the V register.
>> -; CHECK-OPT-NOT: fmov
>> +; CHECK-NOT: fmov
>> ; CHECK: fmov [[COPY_REG2:x[0-9]+]], d[[REG2]]
>> -; CHECK-NOOPT: fmov d0, [[COPY_REG3]]
>> -; CHECK-OPT-NOT: fmov
>> +; CHECK-NOT: fmov
>> ; CHECK: ins.d v0[1], [[COPY_REG2]]
>> ; CHECK-NEXT: ret
>> ;
>> @@ -24,11 +18,9 @@ define <2 x i64> @bar(<2 x i64> %a, <2 x
>> ; GENERIC: addv[[REG:[0-9]+]].2d, v0.2d, v1.2d
>> ; GENERIC: addd[[REG3:[0-9]+]], d[[REG]], d1
>> ; GENERIC: subd[[REG2:[0-9]+]], d[[REG]], d1
>> -; GENERIC-NOOPT: fmov [[COPY_REG3:x[0-9]+]], d[[REG3]]
>> -; GENERIC-OPT-NOT: fmov
>> +; GENERIC-NOT: fmov
>> ; GENERIC: fmov [[COPY_REG2:x[0-9]+]], d[[REG2]]
>> -; GENERIC-NOOPT: fmov d0, [[COPY_REG3]]
>> -; GENERIC-OPT-NOT: fmov
>> +; GENERIC-NOT: fmov
>> ; GENERIC: ins v0.d[1], [[COPY_REG2]]
>> ; GENERIC-NEXT: ret
>>   %add = add <2 x i64> %a, %b
>>
>> Modified: llvm/trunk/test/CodeGen/AArch64/arm64-zero-cycle-regmov.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/arm64-zero-cycle-regmov.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AArch64/arm64-zero-cycle-regmov.ll (original)
>> +++ llvm/trunk/test/CodeGen/AArch64/arm64-zero-cycle-regmov.ll Mon Oct 
>>  2 15:01:37 2017
>> @@ -4,8 +4,10 @@
>> define i32 @t(i32 %a, i32 %b, i32 %c, i32 %d) nounwind ssp {
>> entry:
>> ; CHECK-LABEL: t:
>> -; CHECK: mov x0, [[REG1:x[0-9]+]]
>> -; CHECK: mov x1, [[REG2:x[0-9]+]]
>> +; CHECK: mov [[REG2:x[0-9]+]], x3
>> +; CHECK: mov [[REG1:x[0-9]+]], x2
>> +; CHECK: mov x0, x2
>> +; CHECK: mov x1, x3
>> ; CHECK: bl _foo
>> ; CHECK: mov x0, [[REG1]]
>> ; CHECK: mov x1, [[REG2]]
>>
>> Modified: llvm/trunk/test/CodeGen/AArch64/f16-instructions.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/f16-instructions.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AArch64/f16-instructions.ll (original)
>> +++ llvm/trunk/test/CodeGen/AArch64/f16-instructions.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -489,7 +489,7 @@ else:
>>
>> ; CHECK-COMMON-LABEL: test_phi:
>> ; CHECK-COMMON: mov  x[[PTR:[0-9]+]], x0
>> -; CHECK-COMMON: ldr  h[[AB:[0-9]+]], [x[[PTR]]]
>> +; CHECK-COMMON: ldr  h[[AB:[0-9]+]], [x0]
>> ; CHECK-COMMON: [[LOOP:LBB[0-9_]+]]:
>> ; CHECK-COMMON: mov.16b  v[[R:[0-9]+]], v[[AB]]
>> ; CHECK-COMMON: ldr  h[[AB]], [x[[PTR]]]
>>
>> Modified: llvm/trunk/test/CodeGen/AArch64/flags-multiuse.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/flags-multiuse.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AArch64/flags-multiuse.ll (original)
>> +++ llvm/trunk/test/CodeGen/AArch64/flags-multiuse.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -17,6 +17,9 @@ define i32 @test_multiflag(i32 %n, i32 %
>>   %val = zext i1 %test to i32
>> ; CHECK: cset {{[xw][0-9]+}}, ne
>>
>> +; CHECK: mov [[RHSCOPY:w[0-9]+]], [[RHS]]
>> +; CHECK: mov [[LHSCOPY:w[0-9]+]], [[LHS]]
>> +
>>   store i32 %val, i32* @var
>>
>>   call void @bar()
>> @@ -25,7 +28,7 @@ define i32 @test_multiflag(i32 %n, i32 %
>>   ; Currently, the comparison is emitted again. An MSR/MRS pair would 
>> also be
>>   ; acceptable, but assuming the call preserves NZCV is not.
>>   br i1 %test, label %iftrue, label %iffalse
>> -; CHECK: cmp [[LHS]], [[RHS]]
>> +; CHECK: cmp [[LHSCOPY]], [[RHSCOPY]]
>> ; CHECK: b.eq
>>
>> iftrue:
>>
>> Modified: llvm/trunk/test/CodeGen/AArch64/merge-store-dependency.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/merge-store-dependency.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AArch64/merge-store-dependency.ll (original)
>> +++ llvm/trunk/test/CodeGen/AArch64/merge-store-dependency.ll Mon Oct 
>>  2 15:01:37 2017
>> @@ -8,10 +8,9 @@
>> define void @test(%struct1* %fde, i32 %fd, void (i32, i32, i8*)* 
>> %func, i8* %arg)  {
>> ;CHECK-LABEL: test
>> entry:
>> -; A53: mov [[DATA:w[0-9]+]], w1
>> ; A53: str q{{[0-9]+}}, {{.*}}
>> ; A53: str q{{[0-9]+}}, {{.*}}
>> -; A53: str [[DATA]], {{.*}}
>> +; A53: str w1, {{.*}}
>>
>>   %0 = bitcast %struct1* %fde to i8*
>>   tail call void @llvm.memset.p0i8.i64(i8* %0, i8 0, i64 40, i32 8, i1 
>> false)
>>
>> Modified: llvm/trunk/test/CodeGen/AArch64/neg-imm.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/neg-imm.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AArch64/neg-imm.ll (original)
>> +++ llvm/trunk/test/CodeGen/AArch64/neg-imm.ll Mon Oct  2 15:01:37 2017
>> @@ -7,8 +7,8 @@ declare void @foo(i32)
>> define void @test(i32 %px) {
>> ; CHECK_LABEL: test:
>> ; CHECK_LABEL: %entry
>> -; CHECK: subs
>> -; CHECK-NEXT: csel
>> +; CHECK: subs [[REG0:w[0-9]+]],
>> +; CHECK: csel {{w[0-9]+}}, wzr, [[REG0]]
>> entry:
>>   %sub = add nsw i32 %px, -1
>>   %cmp = icmp slt i32 %px, 1
>>
>> Modified: llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll 
>> (original)
>> +++ llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll Mon 
>> Oct  2 15:01:37 2017
>> @@ -547,16 +547,16 @@ define void @func_use_every_sgpr_input_c
>> ; GCN: s_mov_b32 s5, s32
>> ; GCN: s_add_u32 s32, s32, 0x300
>>
>> -; GCN-DAG: s_mov_b32 [[SAVE_X:s[0-9]+]], s14
>> -; GCN-DAG: s_mov_b32 [[SAVE_Y:s[0-9]+]], s15
>> -; GCN-DAG: s_mov_b32 [[SAVE_Z:s[0-9]+]], s16
>> +; GCN-DAG: s_mov_b32 [[SAVE_X:s[0-57-9][0-9]*]], s14
>> +; GCN-DAG: s_mov_b32 [[SAVE_Y:s[0-68-9][0-9]*]], s15
>> +; GCN-DAG: s_mov_b32 [[SAVE_Z:s[0-79][0-9]*]], s16
>> ; GCN-DAG: s_mov_b64 {{s\[[0-9]+:[0-9]+\]}}, s[6:7]
>> ; GCN-DAG: s_mov_b64 {{s\[[0-9]+:[0-9]+\]}}, s[8:9]
>> ; GCN-DAG: s_mov_b64 {{s\[[0-9]+:[0-9]+\]}}, s[10:11]
>>
>> -; GCN-DAG: s_mov_b32 s6, [[SAVE_X]]
>> -; GCN-DAG: s_mov_b32 s7, [[SAVE_Y]]
>> -; GCN-DAG: s_mov_b32 s8, [[SAVE_Z]]
>> +; GCN-DAG: s_mov_b32 s6, s14
>> +; GCN-DAG: s_mov_b32 s7, s15
>> +; GCN-DAG: s_mov_b32 s8, s16
>> ; GCN: s_swappc_b64
>>
>> ; GCN: buffer_store_dword v{{[0-9]+}}, off, s[0:3], s5 offset:4
>>
>> Modified: llvm/trunk/test/CodeGen/AMDGPU/mad-mix.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/mad-mix.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AMDGPU/mad-mix.ll (original)
>> +++ llvm/trunk/test/CodeGen/AMDGPU/mad-mix.ll Mon Oct  2 15:01:37 2017
>> @@ -51,7 +51,7 @@ define float @v_mad_mix_f32_f16hi_f16hi_
>>
>> ; GCN-LABEL: {{^}}v_mad_mix_v2f32:
>> ; GFX9: v_mov_b32_e32 v3, v1
>> -; GFX9-NEXT: v_mad_mix_f32 v1, v0, v3, v2 op_sel:[1,1,1]
>> +; GFX9-NEXT: v_mad_mix_f32 v1, v0, v1, v2 op_sel:[1,1,1]
>> ; GFX9-NEXT: v_mad_mix_f32 v0, v0, v3, v2
>>
>> ; CIVI: v_mac_f32
>> @@ -66,7 +66,7 @@ define <2 x float> @v_mad_mix_v2f32(<2 x
>> ; GCN-LABEL: {{^}}v_mad_mix_v2f32_shuffle:
>> ; GCN: s_waitcnt
>> ; GFX9-NEXT: v_mov_b32_e32 v3, v1
>> -; GFX9-NEXT: v_mad_mix_f32 v1, v0, v3, v2 op_sel:[0,1,1]
>> +; GFX9-NEXT: v_mad_mix_f32 v1, v0, v1, v2 op_sel:[0,1,1]
>> ; GFX9-NEXT: v_mad_mix_f32 v0, v0, v3, v2 op_sel:[1,0,1]
>> ; GFX9-NEXT: s_setpc_b64
>>
>> @@ -246,7 +246,7 @@ define float @v_mad_mix_f32_f16lo_f16lo_
>> ; GCN-LABEL: {{^}}v_mad_mix_v2f32_f32imm1:
>> ; GFX9: v_mov_b32_e32 v2, v1
>> ; GFX9: v_mov_b32_e32 v3, 1.0
>> -; GFX9: v_mad_mix_f32 v1, v0, v2, v3 op_sel:[1,1,0] op_sel_hi:[1,1,0] 
>> ; encoding
>> +; GFX9: v_mad_mix_f32 v1, v0, v1, v3 op_sel:[1,1,0] op_sel_hi:[1,1,0] 
>> ; encoding
>> ; GFX9: v_mad_mix_f32 v0, v0, v2, v3 op_sel_hi:[1,1,0] ; encoding
>> define <2 x float> @v_mad_mix_v2f32_f32imm1(<2 x half> %src0, <2 x 
>> half> %src1) #0 {
>>   %src0.ext = fpext <2 x half> %src0 to <2 x float>
>> @@ -258,7 +258,7 @@ define <2 x float> @v_mad_mix_v2f32_f32i
>> ; GCN-LABEL: {{^}}v_mad_mix_v2f32_cvtf16imminv2pi:
>> ; GFX9: v_mov_b32_e32 v2, v1
>> ; GFX9: v_mov_b32_e32 v3, 0x3e230000
>> -; GFX9: v_mad_mix_f32 v1, v0, v2, v3 op_sel:[1,1,0] op_sel_hi:[1,1,0] 
>> ; encoding
>> +; GFX9: v_mad_mix_f32 v1, v0, v1, v3 op_sel:[1,1,0] op_sel_hi:[1,1,0] 
>> ; encoding
>> ; GFX9: v_mad_mix_f32 v0, v0, v2, v3 op_sel_hi:[1,1,0] ; encoding
>> define <2 x float> @v_mad_mix_v2f32_cvtf16imminv2pi(<2 x half> %src0, 
>> <2 x half> %src1) #0 {
>>   %src0.ext = fpext <2 x half> %src0 to <2 x float>
>> @@ -271,7 +271,7 @@ define <2 x float> @v_mad_mix_v2f32_cvtf
>> ; GCN-LABEL: {{^}}v_mad_mix_v2f32_f32imminv2pi:
>> ; GFX9: v_mov_b32_e32 v2, v1
>> ; GFX9: v_mov_b32_e32 v3, 0.15915494
>> -; GFX9: v_mad_mix_f32 v1, v0, v2, v3 op_sel:[1,1,0] op_sel_hi:[1,1,0] 
>> ; encoding
>> +; GFX9: v_mad_mix_f32 v1, v0, v1, v3 op_sel:[1,1,0] op_sel_hi:[1,1,0] 
>> ; encoding
>> ; GFX9: v_mad_mix_f32 v0, v0, v2, v3 op_sel_hi:[1,1,0] ; encoding
>> define <2 x float> @v_mad_mix_v2f32_f32imminv2pi(<2 x half> %src0, <2 
>> x half> %src1) #0 {
>>   %src0.ext = fpext <2 x half> %src0 to <2 x float>
>>
>> Modified: llvm/trunk/test/CodeGen/AMDGPU/ret.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/ret.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/AMDGPU/ret.ll (original)
>> +++ llvm/trunk/test/CodeGen/AMDGPU/ret.ll Mon Oct  2 15:01:37 2017
>> @@ -2,10 +2,10 @@
>> ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | 
>> FileCheck -check-prefix=GCN %s
>>
>> ; GCN-LABEL: {{^}}vgpr:
>> -; GCN: v_mov_b32_e32 v1, v0
>> -; GCN-DAG: v_add_f32_e32 v0, 1.0, v1
>> -; GCN-DAG: exp mrt0 v1, v1, v1, v1 done vm
>> +; GCN-DAG: v_mov_b32_e32 v1, v0
>> +; GCN-DAG: exp mrt0 v0, v0, v0, v0 done vm
>> ; GCN: s_waitcnt expcnt(0)
>> +; GCN: v_add_f32_e32 v0, 1.0, v0
>> ; GCN-NOT: s_endpgm
>> define amdgpu_vs { float, float } @vgpr([9 x <16 x i8>] addrspace(2)* 
>> byval %arg, i32 inreg %arg1, i32 inreg %arg2, float %arg3) #0 {
>> bb:
>> @@ -204,13 +204,13 @@ bb:
>> }
>>
>> ; GCN-LABEL: {{^}}both:
>> -; GCN: v_mov_b32_e32 v1, v0
>> -; GCN-DAG: exp mrt0 v1, v1, v1, v1 done vm
>> -; GCN-DAG: v_add_f32_e32 v0, 1.0, v1
>> -; GCN-DAG: s_add_i32 s0, s3, 2
>> +; GCN-DAG: exp mrt0 v0, v0, v0, v0 done vm
>> +; GCN-DAG: v_mov_b32_e32 v1, v0
>> ; GCN-DAG: s_mov_b32 s1, s2
>> -; GCN: s_mov_b32 s2, s3
>> ; GCN: s_waitcnt expcnt(0)
>> +; GCN: v_add_f32_e32 v0, 1.0, v0
>> +; GCN-DAG: s_add_i32 s0, s3, 2
>> +; GCN-DAG: s_mov_b32 s2, s3
>> ; GCN-NOT: s_endpgm
>> define amdgpu_vs { float, i32, float, i32, i32 } @both([9 x <16 x i8>] 
>> addrspace(2)* byval %arg, i32 inreg %arg1, i32 inreg %arg2, float 
>> %arg3) #0 {
>> bb:
>>
>> Modified: llvm/trunk/test/CodeGen/ARM/atomic-op.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/atomic-op.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/ARM/atomic-op.ll (original)
>> +++ llvm/trunk/test/CodeGen/ARM/atomic-op.ll Mon Oct  2 15:01:37 2017
>> @@ -287,7 +287,8 @@ define i32 @test_cmpxchg_fail_order(i32
>>
>>   %pair = cmpxchg i32* %addr, i32 %desired, i32 %new seq_cst monotonic
>>   %oldval = extractvalue { i32, i1 } %pair, 0
>> -; CHECK-ARMV7:     ldrex   [[OLDVAL:r[0-9]+]], [r[[ADDR:[0-9]+]]]
>> +; CHECK-ARMV7:     mov     r[[ADDR:[0-9]+]], r0
>> +; CHECK-ARMV7:     ldrex   [[OLDVAL:r[0-9]+]], [r0]
>> ; CHECK-ARMV7:     cmp     [[OLDVAL]], r1
>> ; CHECK-ARMV7:     bne     [[FAIL_BB:\.?LBB[0-9]+_[0-9]+]]
>> ; CHECK-ARMV7:     dmb ish
>> @@ -305,7 +306,8 @@ define i32 @test_cmpxchg_fail_order(i32
>> ; CHECK-ARMV7:     dmb     ish
>> ; CHECK-ARMV7:     bx      lr
>>
>> -; CHECK-T2:     ldrex   [[OLDVAL:r[0-9]+]], [r[[ADDR:[0-9]+]]]
>> +; CHECK-T2:     mov     r[[ADDR:[0-9]+]], r0
>> +; CHECK-T2:     ldrex   [[OLDVAL:r[0-9]+]], [r0]
>> ; CHECK-T2:     cmp     [[OLDVAL]], r1
>> ; CHECK-T2:     bne     [[FAIL_BB:\.?LBB.*]]
>> ; CHECK-T2:     dmb ish
>>
>> Modified: llvm/trunk/test/CodeGen/ARM/intrinsics-overflow.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/intrinsics-overflow.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/ARM/intrinsics-overflow.ll (original)
>> +++ llvm/trunk/test/CodeGen/ARM/intrinsics-overflow.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -39,7 +39,7 @@ define i32 @sadd_overflow(i32 %a, i32 %b
>>   ; ARM: movvc r[[R1]], #0
>>
>>   ; THUMBV6: mov  r[[R2:[0-9]+]], r[[R0:[0-9]+]]
>> -  ; THUMBV6: adds r[[R3:[0-9]+]], r[[R2]], r[[R1:[0-9]+]]
>> +  ; THUMBV6: adds r[[R3:[0-9]+]], r[[R0]], r[[R1:[0-9]+]]
>>   ; THUMBV6: movs r[[R0]], #0
>>   ; THUMBV6: movs r[[R1]], #1
>>   ; THUMBV6: cmp  r[[R3]], r[[R2]]
>>
>> Modified: llvm/trunk/test/CodeGen/ARM/swifterror.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/swifterror.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/ARM/swifterror.ll (original)
>> +++ llvm/trunk/test/CodeGen/ARM/swifterror.ll Mon Oct  2 15:01:37 2017
>> @@ -182,7 +182,7 @@ define float @foo_loop(%swift_error** sw
>> ; CHECK-APPLE: beq
>> ; CHECK-APPLE: mov r0, #16
>> ; CHECK-APPLE: malloc
>> -; CHECK-APPLE: strb r{{.*}}, [{{.*}}[[ID]], #8]
>> +; CHECK-APPLE: strb r{{.*}}, [r0, #8]
>> ; CHECK-APPLE: ble
>> ; CHECK-APPLE: mov r8, [[ID]]
>>
>>
>> Modified: llvm/trunk/test/CodeGen/Mips/llvm-ir/sub.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Mips/llvm-ir/sub.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/Mips/llvm-ir/sub.ll (original)
>> +++ llvm/trunk/test/CodeGen/Mips/llvm-ir/sub.ll Mon Oct  2 15:01:37 2017
>> @@ -165,7 +165,7 @@ entry:
>> ; MMR3: subu16   $5, $[[T19]], $[[T20]]
>>
>> ; MMR6: move     $[[T0:[0-9]+]], $7
>> -; MMR6: sw       $[[T0]], 8($sp)
>> +; MMR6: sw       $7, 8($sp)
>> ; MMR6: move     $[[T1:[0-9]+]], $5
>> ; MMR6: sw       $4, 12($sp)
>> ; MMR6: lw       $[[T2:[0-9]+]], 48($sp)
>>
>> Modified: llvm/trunk/test/CodeGen/PowerPC/fma-mutate.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/fma-mutate.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/PowerPC/fma-mutate.ll (original)
>> +++ llvm/trunk/test/CodeGen/PowerPC/fma-mutate.ll Mon Oct  2 15:01:37 2017
>> @@ -14,7 +14,8 @@ define double @foo3(double %a) nounwind
>>   ret double %r
>>
>> ; CHECK: @foo3
>> -; CHECK: xsnmsubadp [[REG:[0-9]+]], {{[0-9]+}}, [[REG]]
>> +; CHECK: fmr [[REG:[0-9]+]], [[REG2:[0-9]+]]
>> +; CHECK: xsnmsubadp [[REG]], {{[0-9]+}}, [[REG2]]
>> ; CHECK: xsmaddmdp
>> ; CHECK: xsmaddadp
>> }
>>
>> Modified: llvm/trunk/test/CodeGen/PowerPC/gpr-vsr-spill.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/gpr-vsr-spill.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/PowerPC/gpr-vsr-spill.ll (original)
>> +++ llvm/trunk/test/CodeGen/PowerPC/gpr-vsr-spill.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -16,8 +16,8 @@ if.end:
>>   ret i32 %e.0
>> ; CHECK: @foo
>> ; CHECK: mr [[NEWREG:[0-9]+]], 3
>> +; CHECK: mr [[REG1:[0-9]+]], 4
>> ; CHECK: mtvsrd [[NEWREG2:[0-9]+]], 4
>> -; CHECK: mffprd [[REG1:[0-9]+]], [[NEWREG2]]
>> ; CHECK: add {{[0-9]+}}, [[NEWREG]], [[REG1]]
>> ; CHECK: mffprd [[REG2:[0-9]+]], [[NEWREG2]]
>> ; CHECK: add {{[0-9]+}}, [[REG2]], [[NEWREG]]
>>
>> Modified: llvm/trunk/test/CodeGen/PowerPC/inlineasm-i64-reg.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/inlineasm-i64-reg.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/PowerPC/inlineasm-i64-reg.ll (original)
>> +++ llvm/trunk/test/CodeGen/PowerPC/inlineasm-i64-reg.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -75,7 +75,7 @@ entry:
>>
>> ; CHECK-DAG: mr [[REG:[0-9]+]], 3
>> ; CHECK-DAG: li 0, 1076
>> -; CHECK:     stw [[REG]],
>> +; CHECK-DAG:     stw 3,
>>
>> ; CHECK:     #APP
>> ; CHECK:     sc
>>
>> Modified: llvm/trunk/test/CodeGen/PowerPC/opt-li-add-to-addi.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/opt-li-add-to-addi.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/PowerPC/opt-li-add-to-addi.ll (original)
>> +++ llvm/trunk/test/CodeGen/PowerPC/opt-li-add-to-addi.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -3,7 +3,7 @@
>>
>> define i64 @testOptimizeLiAddToAddi(i64 %a) {
>> ; CHECK-LABEL: testOptimizeLiAddToAddi:
>> -; CHECK:    addi 3, 30, 2444
>> +; CHECK:    addi 3, 3, 2444
>> ; CHECK:    bl callv
>> ; CHECK:    addi 3, 30, 234
>> ; CHECK:    bl call
>>
>> Modified: llvm/trunk/test/CodeGen/PowerPC/tail-dup-layout.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/tail-dup-layout.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/PowerPC/tail-dup-layout.ll (original)
>> +++ llvm/trunk/test/CodeGen/PowerPC/tail-dup-layout.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -23,7 +23,7 @@ target triple = "powerpc64le-grtev4-linu
>> ;CHECK-LABEL: straight_test:
>> ; test1 may have been merged with entry
>> ;CHECK: mr [[TAGREG:[0-9]+]], 3
>> -;CHECK: andi. {{[0-9]+}}, [[TAGREG]], 1
>> +;CHECK: andi. {{[0-9]+}}, [[TAGREG:[0-9]+]], 1
>> ;CHECK-NEXT: bc 12, 1, .[[OPT1LABEL:[_0-9A-Za-z]+]]
>> ;CHECK-NEXT: # %test2
>> ;CHECK-NEXT: rlwinm. {{[0-9]+}}, [[TAGREG]], 0, 30, 30
>>
>> Modified: llvm/trunk/test/CodeGen/SPARC/atomics.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/SPARC/atomics.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/SPARC/atomics.ll (original)
>> +++ llvm/trunk/test/CodeGen/SPARC/atomics.ll Mon Oct  2 15:01:37 2017
>> @@ -235,8 +235,9 @@ entry:
>>
>> ; CHECK-LABEL: test_load_add_i32
>> ; CHECK: membar
>> -; CHECK: add [[V:%[gilo][0-7]]], %o1, [[U:%[gilo][0-7]]]
>> -; CHECK: cas [%o0], [[V]], [[U]]
>> +; CHECK: mov [[U:%[gilo][0-7]]], [[V:%[gilo][0-7]]]
>> +; CHECK: add [[U:%[gilo][0-7]]], %o1, [[V2:%[gilo][0-7]]]
>> +; CHECK: cas [%o0], [[V]], [[V2]]
>> ; CHECK: membar
>> define zeroext i32 @test_load_add_i32(i32* %p, i32 zeroext %v) {
>> entry:
>>
>> Modified: llvm/trunk/test/CodeGen/Thumb/thumb-shrink-wrapping.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Thumb/thumb-shrink-wrapping.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/Thumb/thumb-shrink-wrapping.ll (original)
>> +++ llvm/trunk/test/CodeGen/Thumb/thumb-shrink-wrapping.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -598,7 +598,7 @@ declare void @abort() #0
>> define i32 @b_to_bx(i32 %value) {
>> ; CHECK-LABEL: b_to_bx:
>> ; DISABLE: push {r7, lr}
>> -; CHECK: cmp r1, #49
>> +; CHECK: cmp r0, #49
>> ; CHECK-NEXT: bgt [[ELSE_LABEL:LBB[0-9_]+]]
>> ; ENABLE: push {r7, lr}
>>
>>
>> Modified: llvm/trunk/test/CodeGen/X86/2006-03-01-InstrSchedBug.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2006-03-01-InstrSchedBug.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/2006-03-01-InstrSchedBug.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/2006-03-01-InstrSchedBug.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -7,7 +7,7 @@ define i32 @f(i32 %a, i32 %b) {
>> ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
>> ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
>> ; CHECK-NEXT:    movl %ecx, %edx
>> -; CHECK-NEXT:    imull %edx, %edx
>> +; CHECK-NEXT:    imull %ecx, %edx
>> ; CHECK-NEXT:    imull %eax, %ecx
>> ; CHECK-NEXT:    imull %eax, %eax
>> ; CHECK-NEXT:    addl %edx, %eax
>>
>> Modified: llvm/trunk/test/CodeGen/X86/arg-copy-elide.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/arg-copy-elide.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/arg-copy-elide.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/arg-copy-elide.ll Mon Oct  2 15:01:37 2017
>> @@ -106,7 +106,7 @@ entry:
>> ; CHECK-DAG: movl %edx, %[[r1:[^ ]*]]
>> ; CHECK-DAG: movl 8(%ebp), %[[r2:[^ ]*]]
>> ; CHECK-DAG: movl %[[r2]], 4(%esp)
>> -; CHECK-DAG: movl %[[r1]], (%esp)
>> +; CHECK-DAG: movl %edx, (%esp)
>> ; CHECK: movl %esp, %[[reg:[^ ]*]]
>> ; CHECK: pushl %[[reg]]
>> ; CHECK: calll _addrof_i64
>>
>> Modified: llvm/trunk/test/CodeGen/X86/avg.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avg.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/avg.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/avg.ll Mon Oct  2 15:01:37 2017
>> @@ -407,7 +407,6 @@ define void @avg_v64i8(<64 x i8>* %a, <6
>> ; SSE2-NEXT:    pand %xmm0, %xmm2
>> ; SSE2-NEXT:    packuswb %xmm1, %xmm2
>> ; SSE2-NEXT:    packuswb %xmm10, %xmm2
>> -; SSE2-NEXT:    movdqa %xmm2, %xmm1
>> ; SSE2-NEXT:    psrld $1, %xmm4
>> ; SSE2-NEXT:    psrld $1, %xmm12
>> ; SSE2-NEXT:    pand %xmm0, %xmm12
>> @@ -444,7 +443,7 @@ define void @avg_v64i8(<64 x i8>* %a, <6
>> ; SSE2-NEXT:    movdqu %xmm7, (%rax)
>> ; SSE2-NEXT:    movdqu %xmm11, (%rax)
>> ; SSE2-NEXT:    movdqu %xmm13, (%rax)
>> -; SSE2-NEXT:    movdqu %xmm1, (%rax)
>> +; SSE2-NEXT:    movdqu %xmm2, (%rax)
>> ; SSE2-NEXT:    retq
>> ;
>> ; AVX1-LABEL: avg_v64i8:
>>
>> Modified: llvm/trunk/test/CodeGen/X86/avx-load-store.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-load-store.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/avx-load-store.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/avx-load-store.ll Mon Oct  2 15:01:37 2017
>> @@ -12,11 +12,11 @@ define void @test_256_load(double* nocap
>> ; CHECK-NEXT:    movq %rdx, %r14
>> ; CHECK-NEXT:    movq %rsi, %r15
>> ; CHECK-NEXT:    movq %rdi, %rbx
>> -; CHECK-NEXT:    vmovaps (%rbx), %ymm0
>> +; CHECK-NEXT:    vmovaps (%rdi), %ymm0
>> ; CHECK-NEXT:    vmovups %ymm0, {{[0-9]+}}(%rsp) # 32-byte Spill
>> -; CHECK-NEXT:    vmovaps (%r15), %ymm1
>> +; CHECK-NEXT:    vmovaps (%rsi), %ymm1
>> ; CHECK-NEXT:    vmovups %ymm1, {{[0-9]+}}(%rsp) # 32-byte Spill
>> -; CHECK-NEXT:    vmovaps (%r14), %ymm2
>> +; CHECK-NEXT:    vmovaps (%rdx), %ymm2
>> ; CHECK-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
>> ; CHECK-NEXT:    callq dummy
>> ; CHECK-NEXT:    vmovups {{[0-9]+}}(%rsp), %ymm0 # 32-byte Reload
>>
>> Modified: llvm/trunk/test/CodeGen/X86/avx512-bugfix-25270.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx512-bugfix-25270.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/avx512-bugfix-25270.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/avx512-bugfix-25270.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -9,10 +9,10 @@ define void @bar__512(<16 x i32>* %var)
>> ; CHECK-NEXT:    pushq %rbx
>> ; CHECK-NEXT:    subq $112, %rsp
>> ; CHECK-NEXT:    movq %rdi, %rbx
>> -; CHECK-NEXT:    vmovups (%rbx), %zmm0
>> +; CHECK-NEXT:    vmovups (%rdi), %zmm0
>> ; CHECK-NEXT:    vmovups %zmm0, (%rsp) ## 64-byte Spill
>> ; CHECK-NEXT:    vbroadcastss {{.*}}(%rip), %zmm1
>> -; CHECK-NEXT:    vmovaps %zmm1, (%rbx)
>> +; CHECK-NEXT:    vmovaps %zmm1, (%rdi)
>> ; CHECK-NEXT:    callq _Print__512
>> ; CHECK-NEXT:    vmovups (%rsp), %zmm0 ## 64-byte Reload
>> ; CHECK-NEXT:    callq _Print__512
>>
>> Modified: llvm/trunk/test/CodeGen/X86/avx512-calling-conv.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx512-calling-conv.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/avx512-calling-conv.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/avx512-calling-conv.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -466,7 +466,7 @@ define i32 @test12(i32 %a1, i32 %a2, i32
>> ; KNL_X32-NEXT:    movl %edi, (%esp)
>> ; KNL_X32-NEXT:    calll _test11
>> ; KNL_X32-NEXT:    movl %eax, %ebx
>> -; KNL_X32-NEXT:    movzbl %bl, %eax
>> +; KNL_X32-NEXT:    movzbl %al, %eax
>> ; KNL_X32-NEXT:    movl %eax, {{[0-9]+}}(%esp)
>> ; KNL_X32-NEXT:    movl %esi, {{[0-9]+}}(%esp)
>> ; KNL_X32-NEXT:    movl %edi, (%esp)
>>
>> Modified: llvm/trunk/test/CodeGen/X86/avx512-mask-op.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx512-mask-op.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/avx512-mask-op.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/avx512-mask-op.ll Mon Oct  2 15:01:37 2017
>> @@ -1175,7 +1175,6 @@ define <8 x i1> @test18(i8 %a, i16 %y) {
>> ; KNL-NEXT:    kmovw %esi, %k0
>> ; KNL-NEXT:    kshiftlw $7, %k0, %k2
>> ; KNL-NEXT:    kshiftrw $15, %k2, %k2
>> -; KNL-NEXT:    kmovw %k2, %eax
>> ; KNL-NEXT:    kshiftlw $6, %k0, %k0
>> ; KNL-NEXT:    kshiftrw $15, %k0, %k0
>> ; KNL-NEXT:    kmovw %k0, %ecx
>> @@ -1188,8 +1187,7 @@ define <8 x i1> @test18(i8 %a, i16 %y) {
>> ; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k0
>> ; KNL-NEXT:    kshiftlw $1, %k0, %k0
>> ; KNL-NEXT:    kshiftrw $1, %k0, %k0
>> -; KNL-NEXT:    kmovw %eax, %k1
>> -; KNL-NEXT:    kshiftlw $7, %k1, %k1
>> +; KNL-NEXT:    kshiftlw $7, %k2, %k1
>> ; KNL-NEXT:    korw %k1, %k0, %k1
>> ; KNL-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
>> ; KNL-NEXT:    vpmovqw %zmm0, %xmm0
>> @@ -1202,20 +1200,16 @@ define <8 x i1> @test18(i8 %a, i16 %y) {
>> ; SKX-NEXT:    kmovd %esi, %k1
>> ; SKX-NEXT:    kshiftlw $7, %k1, %k2
>> ; SKX-NEXT:    kshiftrw $15, %k2, %k2
>> -; SKX-NEXT:    kmovd %k2, %eax
>> ; SKX-NEXT:    kshiftlw $6, %k1, %k1
>> ; SKX-NEXT:    kshiftrw $15, %k1, %k1
>> -; SKX-NEXT:    kmovd %k1, %ecx
>> ; SKX-NEXT:    vpmovm2q %k0, %zmm0
>> -; SKX-NEXT:    kmovd %ecx, %k0
>> -; SKX-NEXT:    vpmovm2q %k0, %zmm1
>> +; SKX-NEXT:    vpmovm2q %k1, %zmm1
>> ; SKX-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,1,2,3,4,5,8,7]
>> ; SKX-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
>> ; SKX-NEXT:    vpmovq2m %zmm2, %k0
>> ; SKX-NEXT:    kshiftlb $1, %k0, %k0
>> ; SKX-NEXT:    kshiftrb $1, %k0, %k0
>> -; SKX-NEXT:    kmovd %eax, %k1
>> -; SKX-NEXT:    kshiftlb $7, %k1, %k1
>> +; SKX-NEXT:    kshiftlb $7, %k2, %k1
>> ; SKX-NEXT:    korb %k1, %k0, %k0
>> ; SKX-NEXT:    vpmovm2w %k0, %xmm0
>> ; SKX-NEXT:    vzeroupper
>> @@ -1227,7 +1221,6 @@ define <8 x i1> @test18(i8 %a, i16 %y) {
>> ; AVX512BW-NEXT:    kmovd %esi, %k0
>> ; AVX512BW-NEXT:    kshiftlw $7, %k0, %k2
>> ; AVX512BW-NEXT:    kshiftrw $15, %k2, %k2
>> -; AVX512BW-NEXT:    kmovd %k2, %eax
>> ; AVX512BW-NEXT:    kshiftlw $6, %k0, %k0
>> ; AVX512BW-NEXT:    kshiftrw $15, %k0, %k0
>> ; AVX512BW-NEXT:    kmovd %k0, %ecx
>> @@ -1240,8 +1233,7 @@ define <8 x i1> @test18(i8 %a, i16 %y) {
>> ; AVX512BW-NEXT:    vptestmq %zmm0, %zmm0, %k0
>> ; AVX512BW-NEXT:    kshiftlw $1, %k0, %k0
>> ; AVX512BW-NEXT:    kshiftrw $1, %k0, %k0
>> -; AVX512BW-NEXT:    kmovd %eax, %k1
>> -; AVX512BW-NEXT:    kshiftlw $7, %k1, %k1
>> +; AVX512BW-NEXT:    kshiftlw $7, %k2, %k1
>> ; AVX512BW-NEXT:    korw %k1, %k0, %k0
>> ; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
>> ; AVX512BW-NEXT:    ## kill: %XMM0<def> %XMM0<kill> %ZMM0<kill>
>> @@ -1254,20 +1246,16 @@ define <8 x i1> @test18(i8 %a, i16 %y) {
>> ; AVX512DQ-NEXT:    kmovw %esi, %k1
>> ; AVX512DQ-NEXT:    kshiftlw $7, %k1, %k2
>> ; AVX512DQ-NEXT:    kshiftrw $15, %k2, %k2
>> -; AVX512DQ-NEXT:    kmovw %k2, %eax
>> ; AVX512DQ-NEXT:    kshiftlw $6, %k1, %k1
>> ; AVX512DQ-NEXT:    kshiftrw $15, %k1, %k1
>> -; AVX512DQ-NEXT:    kmovw %k1, %ecx
>> ; AVX512DQ-NEXT:    vpmovm2q %k0, %zmm0
>> -; AVX512DQ-NEXT:    kmovw %ecx, %k0
>> -; AVX512DQ-NEXT:    vpmovm2q %k0, %zmm1
>> +; AVX512DQ-NEXT:    vpmovm2q %k1, %zmm1
>> ; AVX512DQ-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,1,2,3,4,5,8,7]
>> ; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
>> ; AVX512DQ-NEXT:    vpmovq2m %zmm2, %k0
>> ; AVX512DQ-NEXT:    kshiftlb $1, %k0, %k0
>> ; AVX512DQ-NEXT:    kshiftrb $1, %k0, %k0
>> -; AVX512DQ-NEXT:    kmovw %eax, %k1
>> -; AVX512DQ-NEXT:    kshiftlb $7, %k1, %k1
>> +; AVX512DQ-NEXT:    kshiftlb $7, %k2, %k1
>> ; AVX512DQ-NEXT:    korb %k1, %k0, %k0
>> ; AVX512DQ-NEXT:    vpmovm2q %k0, %zmm0
>> ; AVX512DQ-NEXT:    vpmovqw %zmm0, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/avx512-schedule.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx512-schedule.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/avx512-schedule.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/avx512-schedule.ll Mon Oct  2 15:01:37 
>> 2017
>> @@ -5973,20 +5973,16 @@ define <8 x i1> @vmov_test18(i8 %a, i16
>> ; CHECK-NEXT:    kmovd %esi, %k1
>> ; CHECK-NEXT:    kshiftlw $7, %k1, %k2
>> ; CHECK-NEXT:    kshiftrw $15, %k2, %k2
>> -; CHECK-NEXT:    kmovd %k2, %eax
>> ; CHECK-NEXT:    kshiftlw $6, %k1, %k1
>> ; CHECK-NEXT:    kshiftrw $15, %k1, %k1
>> -; CHECK-NEXT:    kmovd %k1, %ecx
>> ; CHECK-NEXT:    vpmovm2q %k0, %zmm0
>> -; CHECK-NEXT:    kmovd %ecx, %k0
>> -; CHECK-NEXT:    vpmovm2q %k0, %zmm1
>> +; CHECK-NEXT:    vpmovm2q %k1, %zmm1
>> ; CHECK-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,1,2,3,4,5,8,7] sched: 
>> [5:0.50]
>> ; CHECK-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
>> ; CHECK-NEXT:    vpmovq2m %zmm2, %k0
>> ; CHECK-NEXT:    kshiftlb $1, %k0, %k0
>> ; CHECK-NEXT:    kshiftrb $1, %k0, %k0
>> -; CHECK-NEXT:    kmovd %eax, %k1
>> -; CHECK-NEXT:    kshiftlb $7, %k1, %k1
>> +; CHECK-NEXT:    kshiftlb $7, %k2, %k1
>> ; CHECK-NEXT:    korb %k1, %k0, %k0
>> ; CHECK-NEXT:    vpmovm2w %k0, %xmm0
>> ; CHECK-NEXT:    vzeroupper # sched: [4:1.00]
>>
>> Modified: llvm/trunk/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll Mon Oct 
>>  2 15:01:37 2017
>> @@ -2090,7 +2090,7 @@ define i64 @test_mask_cmp_b_512(<64 x i8
>> ; AVX512F-32-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm2
>> ; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = 
>> zmm2[0,1,2,3],zmm0[4,5,6,7]
>> ; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
>> -; AVX512F-32-NEXT:    movl %esi, %eax
>> +; AVX512F-32-NEXT:    movl %ecx, %eax
>> ; AVX512F-32-NEXT:    shrl $30, %eax
>> ; AVX512F-32-NEXT:    kmovd %eax, %k1
>> ; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
>> @@ -2101,7 +2101,7 @@ define i64 @test_mask_cmp_b_512(<64 x i8
>> ; AVX512F-32-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm1
>> ; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = 
>> zmm1[0,1,2,3],zmm0[4,5,6,7]
>> ; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
>> -; AVX512F-32-NEXT:    movl %esi, %eax
>> +; AVX512F-32-NEXT:    movl %ecx, %eax
>> ; AVX512F-32-NEXT:    shrl $31, %eax
>> ; AVX512F-32-NEXT:    kmovd %eax, %k1
>> ; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
>> @@ -2980,7 +2980,7 @@ define i64 @test_mask_x86_avx512_ucmp_b_
>> ; AVX512F-32-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm2
>> ; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = 
>> zmm2[0,1,2,3],zmm0[4,5,6,7]
>> ; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
>> -; AVX512F-32-NEXT:    movl %esi, %eax
>> +; AVX512F-32-NEXT:    movl %ecx, %eax
>> ; AVX512F-32-NEXT:    shrl $30, %eax
>> ; AVX512F-32-NEXT:    kmovd %eax, %k1
>> ; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
>> @@ -2991,7 +2991,7 @@ define i64 @test_mask_x86_avx512_ucmp_b_
>> ; AVX512F-32-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm1
>> ; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = 
>> zmm1[0,1,2,3],zmm0[4,5,6,7]
>> ; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
>> -; AVX512F-32-NEXT:    movl %esi, %eax
>> +; AVX512F-32-NEXT:    movl %ecx, %eax
>> ; AVX512F-32-NEXT:    shrl $31, %eax
>> ; AVX512F-32-NEXT:    kmovd %eax, %k1
>> ; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/buildvec-insertvec.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/buildvec-insertvec.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/buildvec-insertvec.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/buildvec-insertvec.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -38,7 +38,7 @@ define <4 x float> @test_negative_zero_1
>> ; SSE2-LABEL: test_negative_zero_1:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movaps %xmm0, %xmm1
>> -; SSE2-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1]
>> +; SSE2-NEXT:    movhlps {{.*#+}} xmm1 = xmm0[1],xmm1[1]
>> ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
>> ; SSE2-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
>> ; SSE2-NEXT:    xorps %xmm2, %xmm2
>>
>> Modified: llvm/trunk/test/CodeGen/X86/combine-fcopysign.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/combine-fcopysign.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/combine-fcopysign.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/combine-fcopysign.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -231,8 +231,8 @@ define <4 x double> @combine_vec_fcopysi
>> ; SSE-NEXT:    cvtss2sd %xmm2, %xmm4
>> ; SSE-NEXT:    movshdup {{.*#+}} xmm5 = xmm2[1,1,3,3]
>> ; SSE-NEXT:    movaps %xmm2, %xmm6
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm6 = xmm6[1,1]
>> -; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1,2,3]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm6 = xmm2[1],xmm6[1]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm2[2,3]
>> ; SSE-NEXT:    movaps {{.*#+}} xmm7
>> ; SSE-NEXT:    movaps %xmm0, %xmm2
>> ; SSE-NEXT:    andps %xmm7, %xmm2
>> @@ -247,7 +247,7 @@ define <4 x double> @combine_vec_fcopysi
>> ; SSE-NEXT:    orps %xmm0, %xmm4
>> ; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm4[0]
>> ; SSE-NEXT:    movaps %xmm1, %xmm0
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm0 = xmm1[1],xmm0[1]
>> ; SSE-NEXT:    andps %xmm7, %xmm0
>> ; SSE-NEXT:    cvtss2sd %xmm3, %xmm3
>> ; SSE-NEXT:    andps %xmm8, %xmm3
>> @@ -294,7 +294,7 @@ define <4 x float> @combine_vec_fcopysig
>> ; SSE-NEXT:    orps %xmm6, %xmm1
>> ; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
>> ; SSE-NEXT:    movaps %xmm3, %xmm1
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm3[1],xmm1[1]
>> ; SSE-NEXT:    andps %xmm5, %xmm1
>> ; SSE-NEXT:    xorps %xmm6, %xmm6
>> ; SSE-NEXT:    cvtsd2ss %xmm2, %xmm6
>>
>> Modified: llvm/trunk/test/CodeGen/X86/combine-shl.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/combine-shl.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/combine-shl.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/combine-shl.ll Mon Oct  2 15:01:37 2017
>> @@ -194,7 +194,7 @@ define <8 x i32> @combine_vec_shl_ext_sh
>> ; SSE-LABEL: combine_vec_shl_ext_shl0:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE-NEXT:    pmovzxwd {{.*#+}} xmm0 = 
>> xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
>> +; SSE-NEXT:    pmovzxwd {{.*#+}} xmm0 = 
>> xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
>> ; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = 
>> xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
>> ; SSE-NEXT:    pslld $20, %xmm1
>> ; SSE-NEXT:    pslld $20, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/complex-fastmath.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/complex-fastmath.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/complex-fastmath.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/complex-fastmath.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -14,7 +14,7 @@ define <2 x float> @complex_square_f32(<
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
>> ; SSE-NEXT:    movaps %xmm0, %xmm2
>> -; SSE-NEXT:    addss %xmm2, %xmm2
>> +; SSE-NEXT:    addss %xmm0, %xmm2
>> ; SSE-NEXT:    mulss %xmm1, %xmm2
>> ; SSE-NEXT:    mulss %xmm0, %xmm0
>> ; SSE-NEXT:    mulss %xmm1, %xmm1
>> @@ -58,9 +58,9 @@ define <2 x double> @complex_square_f64(
>> ; SSE-LABEL: complex_square_f64:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movaps %xmm0, %xmm1
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm0[1],xmm1[1]
>> ; SSE-NEXT:    movaps %xmm0, %xmm2
>> -; SSE-NEXT:    addsd %xmm2, %xmm2
>> +; SSE-NEXT:    addsd %xmm0, %xmm2
>> ; SSE-NEXT:    mulsd %xmm1, %xmm2
>> ; SSE-NEXT:    mulsd %xmm0, %xmm0
>> ; SSE-NEXT:    mulsd %xmm1, %xmm1
>> @@ -161,9 +161,9 @@ define <2 x double> @complex_mul_f64(<2
>> ; SSE-LABEL: complex_mul_f64:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movaps %xmm0, %xmm2
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm2[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm0[1],xmm2[1]
>> ; SSE-NEXT:    movaps %xmm1, %xmm3
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm3 = xmm3[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm3 = xmm1[1],xmm3[1]
>> ; SSE-NEXT:    movaps %xmm3, %xmm4
>> ; SSE-NEXT:    mulsd %xmm0, %xmm4
>> ; SSE-NEXT:    mulsd %xmm1, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/divide-by-constant.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/divide-by-constant.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/divide-by-constant.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/divide-by-constant.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -318,7 +318,7 @@ define i64 @PR23590(i64 %x) nounwind {
>> ; X64:       # BB#0: # %entry
>> ; X64-NEXT:    movq %rdi, %rcx
>> ; X64-NEXT:    movabsq $6120523590596543007, %rdx # imm = 
>> 0x54F077C718E7C21F
>> -; X64-NEXT:    movq %rcx, %rax
>> +; X64-NEXT:    movq %rdi, %rax
>> ; X64-NEXT:    mulq %rdx
>> ; X64-NEXT:    shrq $12, %rdx
>> ; X64-NEXT:    imulq $12345, %rdx, %rax # imm = 0x3039
>>
>> Modified: llvm/trunk/test/CodeGen/X86/fmaxnum.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fmaxnum.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/fmaxnum.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/fmaxnum.ll Mon Oct  2 15:01:37 2017
>> @@ -18,7 +18,7 @@ declare <8 x double> @llvm.maxnum.v8f64(
>>
>> ; CHECK-LABEL: @test_fmaxf
>> ; SSE:         movaps %xmm0, %xmm2
>> -; SSE-NEXT:    cmpunordss %xmm2, %xmm2
>> +; SSE-NEXT:    cmpunordss %xmm0, %xmm2
>> ; SSE-NEXT:    movaps %xmm2, %xmm3
>> ; SSE-NEXT:    andps %xmm1, %xmm3
>> ; SSE-NEXT:    maxss %xmm0, %xmm1
>> @@ -47,7 +47,7 @@ define float @test_fmaxf_minsize(float %
>>
>> ; CHECK-LABEL: @test_fmax
>> ; SSE:         movapd %xmm0, %xmm2
>> -; SSE-NEXT:    cmpunordsd %xmm2, %xmm2
>> +; SSE-NEXT:    cmpunordsd %xmm0, %xmm2
>> ; SSE-NEXT:    movapd %xmm2, %xmm3
>> ; SSE-NEXT:    andpd %xmm1, %xmm3
>> ; SSE-NEXT:    maxsd %xmm0, %xmm1
>> @@ -74,7 +74,7 @@ define x86_fp80 @test_fmaxl(x86_fp80 %x,
>>
>> ; CHECK-LABEL: @test_intrinsic_fmaxf
>> ; SSE:         movaps %xmm0, %xmm2
>> -; SSE-NEXT:    cmpunordss %xmm2, %xmm2
>> +; SSE-NEXT:    cmpunordss %xmm0, %xmm2
>> ; SSE-NEXT:    movaps %xmm2, %xmm3
>> ; SSE-NEXT:    andps %xmm1, %xmm3
>> ; SSE-NEXT:    maxss %xmm0, %xmm1
>> @@ -95,7 +95,7 @@ define float @test_intrinsic_fmaxf(float
>>
>> ; CHECK-LABEL: @test_intrinsic_fmax
>> ; SSE:         movapd %xmm0, %xmm2
>> -; SSE-NEXT:    cmpunordsd %xmm2, %xmm2
>> +; SSE-NEXT:    cmpunordsd %xmm0, %xmm2
>> ; SSE-NEXT:    movapd %xmm2, %xmm3
>> ; SSE-NEXT:    andpd %xmm1, %xmm3
>> ; SSE-NEXT:    maxsd %xmm0, %xmm1
>>
>> Modified: llvm/trunk/test/CodeGen/X86/fmf-flags.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fmf-flags.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/fmf-flags.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/fmf-flags.ll Mon Oct  2 15:01:37 2017
>> @@ -30,7 +30,7 @@ define float @fast_fmuladd_opts(float %a
>> ; X64-LABEL: fast_fmuladd_opts:
>> ; X64:       # BB#0:
>> ; X64-NEXT:    movaps %xmm0, %xmm1
>> -; X64-NEXT:    addss %xmm1, %xmm1
>> +; X64-NEXT:    addss %xmm0, %xmm1
>> ; X64-NEXT:    addss %xmm0, %xmm1
>> ; X64-NEXT:    movaps %xmm1, %xmm0
>> ; X64-NEXT:    retq
>>
>> Modified: llvm/trunk/test/CodeGen/X86/fminnum.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fminnum.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/fminnum.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/fminnum.ll Mon Oct  2 15:01:37 2017
>> @@ -18,7 +18,7 @@ declare <8 x double> @llvm.minnum.v8f64(
>>
>> ; CHECK-LABEL: @test_fminf
>> ; SSE:         movaps %xmm0, %xmm2
>> -; SSE-NEXT:    cmpunordss %xmm2, %xmm2
>> +; SSE-NEXT:    cmpunordss %xmm0, %xmm2
>> ; SSE-NEXT:    movaps %xmm2, %xmm3
>> ; SSE-NEXT:    andps %xmm1, %xmm3
>> ; SSE-NEXT:    minss %xmm0, %xmm1
>> @@ -40,7 +40,7 @@ define float @test_fminf(float %x, float
>>
>> ; CHECK-LABEL: @test_fmin
>> ; SSE:         movapd %xmm0, %xmm2
>> -; SSE-NEXT:    cmpunordsd %xmm2, %xmm2
>> +; SSE-NEXT:    cmpunordsd %xmm0, %xmm2
>> ; SSE-NEXT:    movapd %xmm2, %xmm3
>> ; SSE-NEXT:    andpd %xmm1, %xmm3
>> ; SSE-NEXT:    minsd %xmm0, %xmm1
>> @@ -67,7 +67,7 @@ define x86_fp80 @test_fminl(x86_fp80 %x,
>>
>> ; CHECK-LABEL: @test_intrinsic_fminf
>> ; SSE:         movaps %xmm0, %xmm2
>> -; SSE-NEXT:    cmpunordss %xmm2, %xmm2
>> +; SSE-NEXT:    cmpunordss %xmm0, %xmm2
>> ; SSE-NEXT:    movaps %xmm2, %xmm3
>> ; SSE-NEXT:    andps %xmm1, %xmm3
>> ; SSE-NEXT:    minss %xmm0, %xmm1
>> @@ -87,7 +87,7 @@ define float @test_intrinsic_fminf(float
>>
>> ; CHECK-LABEL: @test_intrinsic_fmin
>> ; SSE:         movapd %xmm0, %xmm2
>> -; SSE-NEXT:    cmpunordsd %xmm2, %xmm2
>> +; SSE-NEXT:    cmpunordsd %xmm0, %xmm2
>> ; SSE-NEXT:    movapd %xmm2, %xmm3
>> ; SSE-NEXT:    andpd %xmm1, %xmm3
>> ; SSE-NEXT:    minsd %xmm0, %xmm1
>>
>> Modified: llvm/trunk/test/CodeGen/X86/fp128-i128.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fp128-i128.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/fp128-i128.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/fp128-i128.ll Mon Oct  2 15:01:37 2017
>> @@ -227,7 +227,7 @@ define fp128 @TestI128_4(fp128 %x) #0 {
>> ; CHECK:       # BB#0: # %entry
>> ; CHECK-NEXT:    subq $40, %rsp
>> ; CHECK-NEXT:    movaps %xmm0, %xmm1
>> -; CHECK-NEXT:    movaps %xmm1, {{[0-9]+}}(%rsp)
>> +; CHECK-NEXT:    movaps %xmm0, {{[0-9]+}}(%rsp)
>> ; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rax
>> ; CHECK-NEXT:    movq %rax, {{[0-9]+}}(%rsp)
>> ; CHECK-NEXT:    movq $0, (%rsp)
>> @@ -275,7 +275,7 @@ define fp128 @acosl(fp128 %x) #0 {
>> ; CHECK:       # BB#0: # %entry
>> ; CHECK-NEXT:    subq $40, %rsp
>> ; CHECK-NEXT:    movaps %xmm0, %xmm1
>> -; CHECK-NEXT:    movaps %xmm1, {{[0-9]+}}(%rsp)
>> +; CHECK-NEXT:    movaps %xmm0, {{[0-9]+}}(%rsp)
>> ; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rax
>> ; CHECK-NEXT:    movq %rax, {{[0-9]+}}(%rsp)
>> ; CHECK-NEXT:    movq $0, (%rsp)
>>
>> Modified: llvm/trunk/test/CodeGen/X86/haddsub-2.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/haddsub-2.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/haddsub-2.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/haddsub-2.ll Mon Oct  2 15:01:37 2017
>> @@ -908,16 +908,16 @@ define <4 x float> @not_a_hsub_2(<4 x fl
>> ; SSE-LABEL: not_a_hsub_2:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movaps %xmm0, %xmm2
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm2[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm0[1],xmm2[1]
>> ; SSE-NEXT:    movaps %xmm0, %xmm3
>> -; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
>> ; SSE-NEXT:    subss %xmm3, %xmm2
>> ; SSE-NEXT:    movshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
>> ; SSE-NEXT:    subss %xmm3, %xmm0
>> ; SSE-NEXT:    movaps %xmm1, %xmm3
>> -; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm1[2,3]
>> ; SSE-NEXT:    movaps %xmm1, %xmm4
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm4 = xmm4[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm4 = xmm1[1],xmm4[1]
>> ; SSE-NEXT:    subss %xmm4, %xmm3
>> ; SSE-NEXT:    movshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
>> ; SSE-NEXT:    subss %xmm4, %xmm1
>> @@ -965,10 +965,10 @@ define <2 x double> @not_a_hsub_3(<2 x d
>> ; SSE-LABEL: not_a_hsub_3:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movaps %xmm1, %xmm2
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm2[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm1[1],xmm2[1]
>> ; SSE-NEXT:    subsd %xmm2, %xmm1
>> ; SSE-NEXT:    movaps %xmm0, %xmm2
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm2[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm0[1],xmm2[1]
>> ; SSE-NEXT:    subsd %xmm0, %xmm2
>> ; SSE-NEXT:    unpcklpd {{.*#+}} xmm2 = xmm2[0],xmm1[0]
>> ; SSE-NEXT:    movapd %xmm2, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/haddsub-undef.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/haddsub-undef.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/haddsub-undef.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/haddsub-undef.ll Mon Oct  2 15:01:37 2017
>> @@ -103,7 +103,7 @@ define <2 x double> @test5_undef(<2 x do
>> ; SSE-LABEL: test5_undef:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movaps %xmm0, %xmm1
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm0[1],xmm1[1]
>> ; SSE-NEXT:    addsd %xmm0, %xmm1
>> ; SSE-NEXT:    movapd %xmm1, %xmm0
>> ; SSE-NEXT:    retq
>> @@ -168,7 +168,7 @@ define <4 x float> @test8_undef(<4 x flo
>> ; SSE-NEXT:    movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
>> ; SSE-NEXT:    addss %xmm0, %xmm1
>> ; SSE-NEXT:    movaps %xmm0, %xmm2
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm2[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm0[1],xmm2[1]
>> ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1,2,3]
>> ; SSE-NEXT:    addss %xmm2, %xmm0
>> ; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
>>
>> Modified: llvm/trunk/test/CodeGen/X86/half.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/half.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/half.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/half.ll Mon Oct  2 15:01:37 2017
>> @@ -386,7 +386,7 @@ define <4 x float> @test_extend32_vec4(<
>> ; CHECK-LIBCALL-NEXT:    pushq %rbx
>> ; CHECK-LIBCALL-NEXT:    subq $48, %rsp
>> ; CHECK-LIBCALL-NEXT:    movq %rdi, %rbx
>> -; CHECK-LIBCALL-NEXT:    movzwl (%rbx), %edi
>> +; CHECK-LIBCALL-NEXT:    movzwl (%rdi), %edi
>> ; CHECK-LIBCALL-NEXT:    callq __gnu_h2f_ieee
>> ; CHECK-LIBCALL-NEXT:    movaps %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill
>> ; CHECK-LIBCALL-NEXT:    movzwl 2(%rbx), %edi
>> @@ -472,7 +472,7 @@ define <4 x double> @test_extend64_vec4(
>> ; CHECK-LIBCALL-NEXT:    pushq %rbx
>> ; CHECK-LIBCALL-NEXT:    subq $16, %rsp
>> ; CHECK-LIBCALL-NEXT:    movq %rdi, %rbx
>> -; CHECK-LIBCALL-NEXT:    movzwl 4(%rbx), %edi
>> +; CHECK-LIBCALL-NEXT:    movzwl 4(%rdi), %edi
>> ; CHECK-LIBCALL-NEXT:    callq __gnu_h2f_ieee
>> ; CHECK-LIBCALL-NEXT:    movss %xmm0, {{[0-9]+}}(%rsp) # 4-byte Spill
>> ; CHECK-LIBCALL-NEXT:    movzwl 6(%rbx), %edi
>> @@ -657,7 +657,7 @@ define void @test_trunc32_vec4(<4 x floa
>> ; CHECK-I686-NEXT:    movaps %xmm0, {{[0-9]+}}(%esp) # 16-byte Spill
>> ; CHECK-I686-NEXT:    movl {{[0-9]+}}(%esp), %ebp
>> ; CHECK-I686-NEXT:    movaps %xmm0, %xmm1
>> -; CHECK-I686-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1,2,3]
>> +; CHECK-I686-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[2,3]
>> ; CHECK-I686-NEXT:    movss %xmm1, (%esp)
>> ; CHECK-I686-NEXT:    calll __gnu_f2h_ieee
>> ; CHECK-I686-NEXT:    movw %ax, %si
>>
>> Modified: llvm/trunk/test/CodeGen/X86/inline-asm-fpstack.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/inline-asm-fpstack.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/inline-asm-fpstack.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/inline-asm-fpstack.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -162,6 +162,7 @@ define void @testPR4459(x86_fp80 %a) {
>> ; CHECK-NEXT:    fstpt (%esp)
>> ; CHECK-NEXT:    calll _ceil
>> ; CHECK-NEXT:    fld %st(0)
>> +; CHECK-NEXT:    fxch %st(1)
>> ; CHECK-NEXT:    ## InlineAsm Start
>> ; CHECK-NEXT:    fistpl %st(0)
>> ; CHECK-NEXT:    ## InlineAsm End
>>
>> Modified: llvm/trunk/test/CodeGen/X86/ipra-local-linkage.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/ipra-local-linkage.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/ipra-local-linkage.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/ipra-local-linkage.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -24,7 +24,7 @@ define void @bar(i32 %X) {
>>   call void @foo()
>>   ; CHECK-LABEL: bar:
>>   ; CHECK: callq foo
>> -  ; CHECK-NEXT: movl  %eax, %r15d
>> +  ; CHECK-NEXT: movl  %edi, %r15d
>>   call void asm sideeffect "movl  $0, %r12d", "{r15}~{r12}"(i32 %X)
>>   ret void
>> }
>>
>> Modified: llvm/trunk/test/CodeGen/X86/localescape.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/localescape.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/localescape.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/localescape.ll Mon Oct  2 15:01:37 2017
>> @@ -27,7 +27,7 @@ define void @print_framealloc_from_fp(i8
>>
>> ; X64-LABEL: print_framealloc_from_fp:
>> ; X64: movq %rcx, %[[parent_fp:[a-z]+]]
>> -; X64: movl .Lalloc_func$frame_escape_0(%[[parent_fp]]), %edx
>> +; X64: movl .Lalloc_func$frame_escape_0(%rcx), %edx
>> ; X64: leaq {{.*}}(%rip), %[[str:[a-z]+]]
>> ; X64: movq %[[str]], %rcx
>> ; X64: callq printf
>>
>> Modified: llvm/trunk/test/CodeGen/X86/machine-cp.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/machine-cp.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/machine-cp.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/machine-cp.ll Mon Oct  2 15:01:37 2017
>> @@ -8,7 +8,7 @@ define i32 @t1(i32 %a, i32 %b) nounwind
>> ; CHECK:       ## BB#0: ## %entry
>> ; CHECK-NEXT:    movl %esi, %edx
>> ; CHECK-NEXT:    movl %edi, %eax
>> -; CHECK-NEXT:    testl %edx, %edx
>> +; CHECK-NEXT:    testl %esi, %esi
>> ; CHECK-NEXT:    je LBB0_1
>> ; CHECK-NEXT:    .p2align 4, 0x90
>> ; CHECK-NEXT:  LBB0_2: ## %while.body
>> @@ -59,7 +59,7 @@ define i32 @t3(i64 %a, i64 %b) nounwind
>> ; CHECK:       ## BB#0: ## %entry
>> ; CHECK-NEXT:    movq %rsi, %rdx
>> ; CHECK-NEXT:    movq %rdi, %rax
>> -; CHECK-NEXT:    testq %rdx, %rdx
>> +; CHECK-NEXT:    testq %rsi, %rsi
>> ; CHECK-NEXT:    je LBB2_1
>> ; CHECK-NEXT:    .p2align 4, 0x90
>> ; CHECK-NEXT:  LBB2_2: ## %while.body
>>
>> Modified: llvm/trunk/test/CodeGen/X86/mul-i1024.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/mul-i1024.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/mul-i1024.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/mul-i1024.ll Mon Oct  2 15:01:37 2017
>> @@ -159,7 +159,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
>> ; X32-NEXT:    pushl %esi
>> ; X32-NEXT:    movl %esi, %ebx
>> -; X32-NEXT:    movl %ebx, {{[0-9]+}}(%esp) # 4-byte Spill
>> +; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
>> ; X32-NEXT:    pushl %edi
>> ; X32-NEXT:    movl %edi, {{[0-9]+}}(%esp) # 4-byte Spill
>> ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
>> @@ -752,7 +752,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    pushl %edi
>> ; X32-NEXT:    movl %ebx, %esi
>> -; X32-NEXT:    pushl %esi
>> +; X32-NEXT:    pushl %ebx
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    pushl {{[0-9]+}}(%esp) # 4-byte Folded Reload
>> @@ -898,7 +898,6 @@ define void @test_1024(i1024* %a, i1024*
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    pushl %edi
>> -; X32-NEXT:    movl %edi, %ebx
>> ; X32-NEXT:    pushl %esi
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    pushl $0
>> @@ -910,7 +909,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X32-NEXT:    leal {{[0-9]+}}(%esp), %eax
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    pushl $0
>> -; X32-NEXT:    pushl %ebx
>> +; X32-NEXT:    pushl %edi
>> ; X32-NEXT:    pushl %esi
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    pushl $0
>> @@ -1365,7 +1364,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    pushl $0
>> ; X32-NEXT:    movl %edi, %ebx
>> -; X32-NEXT:    pushl %ebx
>> +; X32-NEXT:    pushl %edi
>> ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
>> ; X32-NEXT:    pushl %esi
>> ; X32-NEXT:    pushl $0
>> @@ -2442,7 +2441,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
>> ; X32-NEXT:    adcl %edi, %eax
>> ; X32-NEXT:    movl %eax, %esi
>> -; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
>> +; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
>> ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
>> ; X32-NEXT:    addl %eax, {{[0-9]+}}(%esp) # 4-byte Folded Spill
>> ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
>> @@ -4265,7 +4264,6 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq $0, %rbp
>> ; X64-NEXT:    addq %rcx, %rbx
>> ; X64-NEXT:    movq %rbx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> -; X64-NEXT:    movq %rcx, %r11
>> ; X64-NEXT:    adcq %rdi, %rbp
>> ; X64-NEXT:    setb %bl
>> ; X64-NEXT:    movzbl %bl, %ebx
>> @@ -4275,12 +4273,12 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    mulq %r8
>> ; X64-NEXT:    movq %rax, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rdx, {{[0-9]+}}(%rsp) # 8-byte Spill
>> -; X64-NEXT:    movq %r11, %r12
>> -; X64-NEXT:    movq %r11, %r8
>> +; X64-NEXT:    movq %rcx, %r12
>> +; X64-NEXT:    movq %rcx, %r8
>> ; X64-NEXT:    addq %rax, %r12
>> ; X64-NEXT:    movq %rdi, %rax
>> ; X64-NEXT:    movq %rdi, %r9
>> -; X64-NEXT:    movq %r9, (%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %rdi, (%rsp) # 8-byte Spill
>> ; X64-NEXT:    adcq %rdx, %rax
>> ; X64-NEXT:    addq %rbp, %r12
>> ; X64-NEXT:    movq %r12, {{[0-9]+}}(%rsp) # 8-byte Spill
>> @@ -4309,7 +4307,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq %rdx, %rbx
>> ; X64-NEXT:    movq 16(%rsi), %rax
>> ; X64-NEXT:    movq %rsi, %r13
>> -; X64-NEXT:    movq %r13, {{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %rsi, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    mulq %r11
>> ; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> @@ -4322,7 +4320,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq %rbx, %r11
>> ; X64-NEXT:    movq %r8, %rax
>> ; X64-NEXT:    movq %r8, %rbp
>> -; X64-NEXT:    movq %rbp, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %r8, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    addq %rdi, %rax
>> ; X64-NEXT:    movq %r9, %rax
>> ; X64-NEXT:    adcq %rcx, %rax
>> @@ -4334,8 +4332,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %rdx, %rsi
>> ; X64-NEXT:    movq %rax, %rbx
>> ; X64-NEXT:    addq %rdi, %rax
>> -; X64-NEXT:    movq %rdi, %r9
>> -; X64-NEXT:    movq %rsi, %rax
>> +; X64-NEXT:    movq %rdx, %rax
>> ; X64-NEXT:    adcq %rcx, %rax
>> ; X64-NEXT:    movq %rax, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq 32(%r13), %rax
>> @@ -4351,9 +4348,10 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq %rdx, %rax
>> ; X64-NEXT:    movq %rax, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rbp, %rax
>> -; X64-NEXT:    addq %r9, %rax
>> +; X64-NEXT:    addq %rdi, %rax
>> ; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> -; X64-NEXT:    movq %r9, {{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %rdi, %r9
>> +; X64-NEXT:    movq %rdi, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rax # 8-byte Reload
>> ; X64-NEXT:    adcq %r15, %rax
>> ; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> @@ -4371,7 +4369,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    addq %rsi, %r11
>> ; X64-NEXT:    movq %rdx, %rbp
>> ; X64-NEXT:    adcq $0, %rbp
>> -; X64-NEXT:    addq %rcx, %r11
>> +; X64-NEXT:    addq %rbx, %r11
>> ; X64-NEXT:    adcq %rsi, %rbp
>> ; X64-NEXT:    movq %rsi, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    setb %bl
>> @@ -4392,11 +4390,11 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq %rbx, %r10
>> ; X64-NEXT:    movq %rcx, %rdx
>> ; X64-NEXT:    movq %rcx, %r12
>> -; X64-NEXT:    movq %r12, {{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %rcx, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    addq %r9, %rdx
>> ; X64-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %r11, %r8
>> -; X64-NEXT:    adcq %r8, %r15
>> +; X64-NEXT:    adcq %r11, %r15
>> ; X64-NEXT:    movq %r15, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    adcq %rax, %r14
>> ; X64-NEXT:    movq %r14, {{[0-9]+}}(%rsp) # 8-byte Spill
>> @@ -4492,13 +4490,12 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq %rdx, %r12
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rcx # 8-byte Reload
>> ; X64-NEXT:    movq %rcx, %rax
>> -; X64-NEXT:    movq %r10, %rbp
>> -; X64-NEXT:    mulq %rbp
>> +; X64-NEXT:    mulq %r10
>> ; X64-NEXT:    movq %rdx, %rsi
>> ; X64-NEXT:    movq %rax, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rdi # 8-byte Reload
>> ; X64-NEXT:    movq %rdi, %rax
>> -; X64-NEXT:    mulq %rbp
>> +; X64-NEXT:    mulq %r10
>> ; X64-NEXT:    movq %rdx, %rbp
>> ; X64-NEXT:    movq %rax, %rbx
>> ; X64-NEXT:    addq %rsi, %rbx
>> @@ -4525,7 +4522,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq $0, %r15
>> ; X64-NEXT:    adcq $0, %r12
>> ; X64-NEXT:    movq %r10, %rbx
>> -; X64-NEXT:    movq %rbx, %rax
>> +; X64-NEXT:    movq %r10, %rax
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r11 # 8-byte Reload
>> ; X64-NEXT:    mulq %r11
>> ; X64-NEXT:    movq %rdx, %rcx
>> @@ -4542,7 +4539,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %rbx, %rax
>> ; X64-NEXT:    mulq %rcx
>> ; X64-NEXT:    movq %rcx, %rbx
>> -; X64-NEXT:    movq %rbx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %rcx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rdx, %rcx
>> ; X64-NEXT:    movq %rax, %r8
>> ; X64-NEXT:    addq %rbp, %r8
>> @@ -4573,7 +4570,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rcx # 8-byte Reload
>> ; X64-NEXT:    movq %rcx, %rax
>> ; X64-NEXT:    movq %r11, %rsi
>> -; X64-NEXT:    mulq %rsi
>> +; X64-NEXT:    mulq %r11
>> ; X64-NEXT:    movq %rdx, %r11
>> ; X64-NEXT:    movq %rax, %r13
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r12 # 8-byte Reload
>> @@ -4653,13 +4650,12 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq %rdx, %r10
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rcx # 8-byte Reload
>> ; X64-NEXT:    movq %rcx, %rax
>> -; X64-NEXT:    movq %r11, %rbp
>> -; X64-NEXT:    mulq %rbp
>> +; X64-NEXT:    mulq %r11
>> ; X64-NEXT:    movq %rdx, %rdi
>> ; X64-NEXT:    movq %rax, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rsi # 8-byte Reload
>> ; X64-NEXT:    movq %rsi, %rax
>> -; X64-NEXT:    mulq %rbp
>> +; X64-NEXT:    mulq %r11
>> ; X64-NEXT:    movq %rdx, %rbp
>> ; X64-NEXT:    movq %rax, %rbx
>> ; X64-NEXT:    addq %rdi, %rbx
>> @@ -4789,7 +4785,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %rdx, %rsi
>> ; X64-NEXT:    movq %rax, %r14
>> ; X64-NEXT:    movq %r8, %rbp
>> -; X64-NEXT:    movq %rbp, %rax
>> +; X64-NEXT:    movq %r8, %rax
>> ; X64-NEXT:    mulq %rcx
>> ; X64-NEXT:    movq %rcx, %r11
>> ; X64-NEXT:    movq %rdx, %rbx
>> @@ -4849,7 +4845,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq $0, %r9
>> ; X64-NEXT:    adcq $0, %r10
>> ; X64-NEXT:    movq %rbp, %rsi
>> -; X64-NEXT:    movq %rsi, %rax
>> +; X64-NEXT:    movq %rbp, %rax
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rcx # 8-byte Reload
>> ; X64-NEXT:    mulq %rcx
>> ; X64-NEXT:    movq %rdx, %r14
>> @@ -4906,8 +4902,8 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq $0, %r15
>> ; X64-NEXT:    movq %rbp, %rax
>> ; X64-NEXT:    movq %r8, %rdi
>> -; X64-NEXT:    movq %rdi, {{[0-9]+}}(%rsp) # 8-byte Spill
>> -; X64-NEXT:    mulq %rdi
>> +; X64-NEXT:    movq %r8, {{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    mulq %r8
>> ; X64-NEXT:    movq %rdx, %r9
>> ; X64-NEXT:    movq %rax, %r8
>> ; X64-NEXT:    addq %rbx, %r8
>> @@ -4990,13 +4986,12 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %rcx, %r14
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rcx # 8-byte Reload
>> ; X64-NEXT:    movq %rcx, %rax
>> -; X64-NEXT:    movq %r10, %rdi
>> -; X64-NEXT:    mulq %rdi
>> +; X64-NEXT:    mulq %r10
>> ; X64-NEXT:    movq %rdx, %r11
>> ; X64-NEXT:    movq %rax, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rsi # 8-byte Reload
>> ; X64-NEXT:    movq %rsi, %rax
>> -; X64-NEXT:    mulq %rdi
>> +; X64-NEXT:    mulq %r10
>> ; X64-NEXT:    movq %rdx, %rdi
>> ; X64-NEXT:    movq %rax, %rbx
>> ; X64-NEXT:    addq %r11, %rbx
>> @@ -5024,8 +5019,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %r8, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    adcq $0, %r14
>> ; X64-NEXT:    movq %r14, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> -; X64-NEXT:    movq %r13, %rbx
>> -; X64-NEXT:    movq %rbx, %rax
>> +; X64-NEXT:    movq %r13, %rax
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rcx # 8-byte Reload
>> ; X64-NEXT:    mulq %rcx
>> ; X64-NEXT:    movq %rdx, %r8
>> @@ -5038,7 +5032,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %rax, %rcx
>> ; X64-NEXT:    addq %r8, %rcx
>> ; X64-NEXT:    adcq $0, %rsi
>> -; X64-NEXT:    movq %rbx, %rax
>> +; X64-NEXT:    movq %r13, %rax
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %r13 # 8-byte Reload
>> ; X64-NEXT:    mulq %r13
>> ; X64-NEXT:    movq %rdx, %rbx
>> @@ -5072,13 +5066,12 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    setb -{{[0-9]+}}(%rsp) # 1-byte Folded Spill
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rbx # 8-byte Reload
>> ; X64-NEXT:    movq %rbx, %rax
>> -; X64-NEXT:    movq %r10, %rsi
>> -; X64-NEXT:    mulq %rsi
>> +; X64-NEXT:    mulq %r10
>> ; X64-NEXT:    movq %rdx, %rcx
>> ; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r8 # 8-byte Reload
>> ; X64-NEXT:    movq %r8, %rax
>> -; X64-NEXT:    mulq %rsi
>> +; X64-NEXT:    mulq %r10
>> ; X64-NEXT:    movq %rdx, %rsi
>> ; X64-NEXT:    movq %rax, %rdi
>> ; X64-NEXT:    addq %rcx, %rdi
>> @@ -5154,7 +5147,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %r9, %rax
>> ; X64-NEXT:    mulq %rcx
>> ; X64-NEXT:    movq %rcx, %r10
>> -; X64-NEXT:    movq %r10, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %rcx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rdx, %rcx
>> ; X64-NEXT:    movq %rax, %rdi
>> ; X64-NEXT:    addq %rsi, %rdi
>> @@ -5166,16 +5159,16 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %rax, %rbx
>> ; X64-NEXT:    movq %rdx, %r14
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r12 # 8-byte Reload
>> -; X64-NEXT:    addq %rbx, %r12
>> +; X64-NEXT:    addq %rax, %r12
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %r15 # 8-byte Reload
>> -; X64-NEXT:    adcq %r14, %r15
>> +; X64-NEXT:    adcq %rdx, %r15
>> ; X64-NEXT:    addq %rdi, %r12
>> ; X64-NEXT:    adcq %rcx, %r15
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rcx # 8-byte Reload
>> ; X64-NEXT:    movq %rcx, %rax
>> ; X64-NEXT:    movq %r11, %rsi
>> -; X64-NEXT:    movq %rsi, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> -; X64-NEXT:    mulq %rsi
>> +; X64-NEXT:    movq %r11, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    mulq %r11
>> ; X64-NEXT:    movq %rdx, %r11
>> ; X64-NEXT:    movq %rax, {{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %r9 # 8-byte Reload
>> @@ -5239,7 +5232,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rax, %r9
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rbp # 8-byte Reload
>> -; X64-NEXT:    addq %r9, %rbp
>> +; X64-NEXT:    addq %rax, %rbp
>> ; X64-NEXT:    movq {{[0-9]+}}(%rsp), %rax # 8-byte Reload
>> ; X64-NEXT:    adcq %rdx, %rax
>> ; X64-NEXT:    addq %rsi, %rbp
>> @@ -5417,7 +5410,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq 88(%rsi), %rax
>> ; X64-NEXT:    movq %rsi, %r9
>> ; X64-NEXT:    movq %rax, %rsi
>> -; X64-NEXT:    movq %rsi, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    mulq %rcx
>> ; X64-NEXT:    movq %rcx, %r11
>> ; X64-NEXT:    movq %rdx, %rbp
>> @@ -5453,13 +5446,12 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    adcq %r8, %r10
>> ; X64-NEXT:    addq %rbx, %rsi
>> ; X64-NEXT:    adcq %rbp, %r10
>> -; X64-NEXT:    movq %r9, %rdi
>> -; X64-NEXT:    movq 64(%rdi), %r13
>> +; X64-NEXT:    movq 64(%r9), %r13
>> ; X64-NEXT:    movq %r13, %rax
>> ; X64-NEXT:    mulq %r11
>> ; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rdx, %rcx
>> -; X64-NEXT:    movq 72(%rdi), %r9
>> +; X64-NEXT:    movq 72(%r9), %r9
>> ; X64-NEXT:    movq %r9, %rax
>> ; X64-NEXT:    mulq %r11
>> ; X64-NEXT:    movq %rdx, %rbp
>> @@ -5487,8 +5479,8 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    movq %rdx, %r11
>> ; X64-NEXT:    movq %rax, %r15
>> ; X64-NEXT:    movq %r12, %rcx
>> -; X64-NEXT:    addq %r15, %rcx
>> -; X64-NEXT:    adcq %r11, %r8
>> +; X64-NEXT:    addq %rax, %rcx
>> +; X64-NEXT:    adcq %rdx, %r8
>> ; X64-NEXT:    addq %rbp, %rcx
>> ; X64-NEXT:    adcq %rbx, %r8
>> ; X64-NEXT:    addq -{{[0-9]+}}(%rsp), %rcx # 8-byte Folded Reload
>> @@ -5540,14 +5532,13 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    setb %r10b
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rsi # 8-byte Reload
>> ; X64-NEXT:    movq %rsi, %rax
>> -; X64-NEXT:    movq %r8, %rdi
>> -; X64-NEXT:    mulq %rdi
>> +; X64-NEXT:    mulq %r8
>> ; X64-NEXT:    movq %rdx, %rcx
>> ; X64-NEXT:    movq %rax, %r9
>> ; X64-NEXT:    movq -{{[0-9]+}}(%rsp), %rbp # 8-byte Reload
>> ; X64-NEXT:    movq %rbp, %rax
>> -; X64-NEXT:    mulq %rdi
>> -; X64-NEXT:    movq %rdi, %r12
>> +; X64-NEXT:    mulq %r8
>> +; X64-NEXT:    movq %r8, %r12
>> ; X64-NEXT:    movq %rdx, %rdi
>> ; X64-NEXT:    movq %rax, %rbx
>> ; X64-NEXT:    addq %rcx, %rbx
>> @@ -5586,7 +5577,7 @@ define void @test_1024(i1024* %a, i1024*
>> ; X64-NEXT:    imulq %rcx, %rdi
>> ; X64-NEXT:    movq %rcx, %rax
>> ; X64-NEXT:    movq %r12, %rsi
>> -; X64-NEXT:    mulq %rsi
>> +; X64-NEXT:    mulq %r12
>> ; X64-NEXT:    movq %rax, %r9
>> ; X64-NEXT:    addq %rdi, %rdx
>> ; X64-NEXT:    movq 104(%rbp), %r8
>>
>> Modified: llvm/trunk/test/CodeGen/X86/mul-i512.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/mul-i512.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/mul-i512.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/mul-i512.ll Mon Oct  2 15:01:37 2017
>> @@ -909,7 +909,7 @@ define void @test_512(i512* %a, i512* %b
>> ; X64-NEXT:    movq 8(%rsi), %rbp
>> ; X64-NEXT:    movq %r15, %rax
>> ; X64-NEXT:    movq %rdx, %rsi
>> -; X64-NEXT:    mulq %rsi
>> +; X64-NEXT:    mulq %rdx
>> ; X64-NEXT:    movq %rdx, %r9
>> ; X64-NEXT:    movq %rax, %r8
>> ; X64-NEXT:    movq %r11, %rax
>> @@ -932,23 +932,24 @@ define void @test_512(i512* %a, i512* %b
>> ; X64-NEXT:    movq %r11, %rax
>> ; X64-NEXT:    mulq %rbp
>> ; X64-NEXT:    movq %rbp, %r14
>> -; X64-NEXT:    movq %r14, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %rbp, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rdx, %rsi
>> ; X64-NEXT:    movq %rax, %rbp
>> ; X64-NEXT:    addq %rcx, %rbp
>> ; X64-NEXT:    adcq %rbx, %rsi
>> ; X64-NEXT:    xorl %ecx, %ecx
>> ; X64-NEXT:    movq %r10, %rbx
>> -; X64-NEXT:    movq %rbx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> -; X64-NEXT:    movq %rbx, %rax
>> +; X64-NEXT:    movq %r10, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %r10, %rax
>> ; X64-NEXT:    mulq %rcx
>> ; X64-NEXT:    movq %rdx, %r13
>> ; X64-NEXT:    movq %rax, %r10
>> ; X64-NEXT:    movq %r15, %rax
>> ; X64-NEXT:    mulq %rcx
>> ; X64-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    # kill: %RAX<kill>
>> +; X64-NEXT:    movq %rax, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rax, %r15
>> -; X64-NEXT:    movq %r15, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    addq %r10, %r15
>> ; X64-NEXT:    adcq %r13, %rdx
>> ; X64-NEXT:    addq %rbp, %r15
>> @@ -987,8 +988,8 @@ define void @test_512(i512* %a, i512* %b
>> ; X64-NEXT:    mulq %rdx
>> ; X64-NEXT:    movq %rdx, %r14
>> ; X64-NEXT:    movq %rax, %r11
>> -; X64-NEXT:    addq %r11, %r10
>> -; X64-NEXT:    adcq %r14, %r13
>> +; X64-NEXT:    addq %rax, %r10
>> +; X64-NEXT:    adcq %rdx, %r13
>> ; X64-NEXT:    addq %rbp, %r10
>> ; X64-NEXT:    adcq %rsi, %r13
>> ; X64-NEXT:    addq %r8, %r10
>> @@ -1000,7 +1001,7 @@ define void @test_512(i512* %a, i512* %b
>> ; X64-NEXT:    movq 16(%rsi), %r8
>> ; X64-NEXT:    movq %rcx, %rax
>> ; X64-NEXT:    movq %rcx, %r9
>> -; X64-NEXT:    movq %r9, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> +; X64-NEXT:    movq %rcx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    mulq %r8
>> ; X64-NEXT:    movq %rdx, %rdi
>> ; X64-NEXT:    movq %rax, %r12
>> @@ -1031,7 +1032,7 @@ define void @test_512(i512* %a, i512* %b
>> ; X64-NEXT:    mulq %rcx
>> ; X64-NEXT:    movq %rdx, -{{[0-9]+}}(%rsp) # 8-byte Spill
>> ; X64-NEXT:    movq %rax, %rbp
>> -; X64-NEXT:    addq %rbp, %r11
>> +; X64-NEXT:    addq %rax, %r11
>> ; X64-NEXT:    adcq %rdx, %r14
>> ; X64-NEXT:    addq %r9, %r11
>> ; X64-NEXT:    adcq %rbx, %r14
>>
>> Modified: llvm/trunk/test/CodeGen/X86/mul128.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/mul128.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/mul128.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/mul128.ll Mon Oct  2 15:01:37 2017
>> @@ -7,7 +7,7 @@ define i128 @foo(i128 %t, i128 %u) {
>> ; X64-NEXT:    movq %rdx, %r8
>> ; X64-NEXT:    imulq %rdi, %rcx
>> ; X64-NEXT:    movq %rdi, %rax
>> -; X64-NEXT:    mulq %r8
>> +; X64-NEXT:    mulq %rdx
>> ; X64-NEXT:    addq %rcx, %rdx
>> ; X64-NEXT:    imulq %r8, %rsi
>> ; X64-NEXT:    addq %rsi, %rdx
>>
>> Modified: llvm/trunk/test/CodeGen/X86/mulvi32.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/mulvi32.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/mulvi32.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/mulvi32.ll Mon Oct  2 15:01:37 2017
>> @@ -292,7 +292,7 @@ define <4 x i64> @_mul4xi32toi64b(<4 x i
>> ; SSE-LABEL: _mul4xi32toi64b:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movdqa %xmm0, %xmm2
>> -; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
>> +; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
>> ; SSE-NEXT:    pmuludq %xmm1, %xmm2
>> ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
>> ; SSE-NEXT:    pmuludq %xmm0, %xmm1
>>
>> Modified: llvm/trunk/test/CodeGen/X86/pmul.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/pmul.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/pmul.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/pmul.ll Mon Oct  2 15:01:37 2017
>> @@ -9,7 +9,7 @@ define <16 x i8> @mul_v16i8c(<16 x i8> %
>> ; SSE2-LABEL: mul_v16i8c:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = 
>> xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = 
>> xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
>> ; SSE2-NEXT:    psraw $8, %xmm1
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [117,117,117,117,117,117,117,117]
>> ; SSE2-NEXT:    pmullw %xmm2, %xmm1
>> @@ -143,10 +143,10 @@ define <16 x i8> @mul_v16i8(<16 x i8> %i
>> ; SSE2-LABEL: mul_v16i8:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
>> ; SSE2-NEXT:    psraw $8, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm3
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = 
>> xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = 
>> xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
>> ; SSE2-NEXT:    psraw $8, %xmm3
>> ; SSE2-NEXT:    pmullw %xmm2, %xmm3
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
>> @@ -386,7 +386,7 @@ define <32 x i8> @mul_v32i8c(<32 x i8> %
>> ; SSE2-LABEL: mul_v32i8c:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm2
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
>> ; SSE2-NEXT:    psraw $8, %xmm2
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [117,117,117,117,117,117,117,117]
>> ; SSE2-NEXT:    pmullw %xmm3, %xmm2
>> @@ -398,7 +398,7 @@ define <32 x i8> @mul_v32i8c(<32 x i8> %
>> ; SSE2-NEXT:    pand %xmm4, %xmm0
>> ; SSE2-NEXT:    packuswb %xmm2, %xmm0
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
>> ; SSE2-NEXT:    psraw $8, %xmm2
>> ; SSE2-NEXT:    pmullw %xmm3, %xmm2
>> ; SSE2-NEXT:    pand %xmm4, %xmm2
>> @@ -567,10 +567,10 @@ define <32 x i8> @mul_v32i8(<32 x i8> %i
>> ; SSE2-LABEL: mul_v32i8:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm4
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = 
>> xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = 
>> xmm4[8],xmm2[8],xmm4[9],xmm2[9],xmm4[10],xmm2[10],xmm4[11],xmm2[11],xmm4[12],xmm2[12],xmm4[13],xmm2[13],xmm4[14],xmm2[14],xmm4[15],xmm2[15]
>> ; SSE2-NEXT:    psraw $8, %xmm4
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm5
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = 
>> xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = 
>> xmm5[8],xmm0[8],xmm5[9],xmm0[9],xmm5[10],xmm0[10],xmm5[11],xmm0[11],xmm5[12],xmm0[12],xmm5[13],xmm0[13],xmm5[14],xmm0[14],xmm5[15],xmm0[15]
>> ; SSE2-NEXT:    psraw $8, %xmm5
>> ; SSE2-NEXT:    pmullw %xmm4, %xmm5
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
>> @@ -583,10 +583,10 @@ define <32 x i8> @mul_v32i8(<32 x i8> %i
>> ; SSE2-NEXT:    pand %xmm4, %xmm0
>> ; SSE2-NEXT:    packuswb %xmm5, %xmm0
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm2
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
>> ; SSE2-NEXT:    psraw $8, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm5
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = 
>> xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = 
>> xmm5[8],xmm1[8],xmm5[9],xmm1[9],xmm5[10],xmm1[10],xmm5[11],xmm1[11],xmm5[12],xmm1[12],xmm5[13],xmm1[13],xmm5[14],xmm1[14],xmm5[15],xmm1[15]
>> ; SSE2-NEXT:    psraw $8, %xmm5
>> ; SSE2-NEXT:    pmullw %xmm2, %xmm5
>> ; SSE2-NEXT:    pand %xmm4, %xmm5
>> @@ -774,7 +774,7 @@ define <64 x i8> @mul_v64i8c(<64 x i8> %
>> ; SSE2-LABEL: mul_v64i8c:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm6
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = 
>> xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = 
>> xmm6[8],xmm0[8],xmm6[9],xmm0[9],xmm6[10],xmm0[10],xmm6[11],xmm0[11],xmm6[12],xmm0[12],xmm6[13],xmm0[13],xmm6[14],xmm0[14],xmm6[15],xmm0[15]
>> ; SSE2-NEXT:    psraw $8, %xmm6
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [117,117,117,117,117,117,117,117]
>> ; SSE2-NEXT:    pmullw %xmm4, %xmm6
>> @@ -786,7 +786,7 @@ define <64 x i8> @mul_v64i8c(<64 x i8> %
>> ; SSE2-NEXT:    pand %xmm5, %xmm0
>> ; SSE2-NEXT:    packuswb %xmm6, %xmm0
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm6
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = 
>> xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = 
>> xmm6[8],xmm1[8],xmm6[9],xmm1[9],xmm6[10],xmm1[10],xmm6[11],xmm1[11],xmm6[12],xmm1[12],xmm6[13],xmm1[13],xmm6[14],xmm1[14],xmm6[15],xmm1[15]
>> ; SSE2-NEXT:    psraw $8, %xmm6
>> ; SSE2-NEXT:    pmullw %xmm4, %xmm6
>> ; SSE2-NEXT:    pand %xmm5, %xmm6
>> @@ -796,7 +796,7 @@ define <64 x i8> @mul_v64i8c(<64 x i8> %
>> ; SSE2-NEXT:    pand %xmm5, %xmm1
>> ; SSE2-NEXT:    packuswb %xmm6, %xmm1
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm6
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = 
>> xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = 
>> xmm6[8],xmm2[8],xmm6[9],xmm2[9],xmm6[10],xmm2[10],xmm6[11],xmm2[11],xmm6[12],xmm2[12],xmm6[13],xmm2[13],xmm6[14],xmm2[14],xmm6[15],xmm2[15]
>> ; SSE2-NEXT:    psraw $8, %xmm6
>> ; SSE2-NEXT:    pmullw %xmm4, %xmm6
>> ; SSE2-NEXT:    pand %xmm5, %xmm6
>> @@ -806,7 +806,7 @@ define <64 x i8> @mul_v64i8c(<64 x i8> %
>> ; SSE2-NEXT:    pand %xmm5, %xmm2
>> ; SSE2-NEXT:    packuswb %xmm6, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm6
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = 
>> xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = 
>> xmm6[8],xmm3[8],xmm6[9],xmm3[9],xmm6[10],xmm3[10],xmm6[11],xmm3[11],xmm6[12],xmm3[12],xmm6[13],xmm3[13],xmm6[14],xmm3[14],xmm6[15],xmm3[15]
>> ; SSE2-NEXT:    psraw $8, %xmm6
>> ; SSE2-NEXT:    pmullw %xmm4, %xmm6
>> ; SSE2-NEXT:    pand %xmm5, %xmm6
>> @@ -821,7 +821,7 @@ define <64 x i8> @mul_v64i8c(<64 x i8> %
>> ; SSE41:       # BB#0: # %entry
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm4
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE41-NEXT:    pmovsxbw %xmm1, %xmm0
>> +; SSE41-NEXT:    pmovsxbw %xmm0, %xmm0
>> ; SSE41-NEXT:    movdqa {{.*#+}} xmm6 = [117,117,117,117,117,117,117,117]
>> ; SSE41-NEXT:    pmullw %xmm6, %xmm0
>> ; SSE41-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255]
>> @@ -939,10 +939,10 @@ define <64 x i8> @mul_v64i8(<64 x i8> %i
>> ; SSE2-LABEL: mul_v64i8:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm4, %xmm8
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm8 = 
>> xmm8[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm8 = 
>> xmm8[8],xmm4[8],xmm8[9],xmm4[9],xmm8[10],xmm4[10],xmm8[11],xmm4[11],xmm8[12],xmm4[12],xmm8[13],xmm4[13],xmm8[14],xmm4[14],xmm8[15],xmm4[15]
>> ; SSE2-NEXT:    psraw $8, %xmm8
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm9
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = 
>> xmm9[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = 
>> xmm9[8],xmm0[8],xmm9[9],xmm0[9],xmm9[10],xmm0[10],xmm9[11],xmm0[11],xmm9[12],xmm0[12],xmm9[13],xmm0[13],xmm9[14],xmm0[14],xmm9[15],xmm0[15]
>> ; SSE2-NEXT:    psraw $8, %xmm9
>> ; SSE2-NEXT:    pmullw %xmm8, %xmm9
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
>> @@ -955,10 +955,10 @@ define <64 x i8> @mul_v64i8(<64 x i8> %i
>> ; SSE2-NEXT:    pand %xmm8, %xmm0
>> ; SSE2-NEXT:    packuswb %xmm9, %xmm0
>> ; SSE2-NEXT:    movdqa %xmm5, %xmm9
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = 
>> xmm9[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = 
>> xmm9[8],xmm5[8],xmm9[9],xmm5[9],xmm9[10],xmm5[10],xmm9[11],xmm5[11],xmm9[12],xmm5[12],xmm9[13],xmm5[13],xmm9[14],xmm5[14],xmm9[15],xmm5[15]
>> ; SSE2-NEXT:    psraw $8, %xmm9
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm4
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = 
>> xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = 
>> xmm4[8],xmm1[8],xmm4[9],xmm1[9],xmm4[10],xmm1[10],xmm4[11],xmm1[11],xmm4[12],xmm1[12],xmm4[13],xmm1[13],xmm4[14],xmm1[14],xmm4[15],xmm1[15]
>> ; SSE2-NEXT:    psraw $8, %xmm4
>> ; SSE2-NEXT:    pmullw %xmm9, %xmm4
>> ; SSE2-NEXT:    pand %xmm8, %xmm4
>> @@ -970,10 +970,10 @@ define <64 x i8> @mul_v64i8(<64 x i8> %i
>> ; SSE2-NEXT:    pand %xmm8, %xmm1
>> ; SSE2-NEXT:    packuswb %xmm4, %xmm1
>> ; SSE2-NEXT:    movdqa %xmm6, %xmm4
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = 
>> xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = 
>> xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
>> ; SSE2-NEXT:    psraw $8, %xmm4
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm5
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = 
>> xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = 
>> xmm5[8],xmm2[8],xmm5[9],xmm2[9],xmm5[10],xmm2[10],xmm5[11],xmm2[11],xmm5[12],xmm2[12],xmm5[13],xmm2[13],xmm5[14],xmm2[14],xmm5[15],xmm2[15]
>> ; SSE2-NEXT:    psraw $8, %xmm5
>> ; SSE2-NEXT:    pmullw %xmm4, %xmm5
>> ; SSE2-NEXT:    pand %xmm8, %xmm5
>> @@ -985,10 +985,10 @@ define <64 x i8> @mul_v64i8(<64 x i8> %i
>> ; SSE2-NEXT:    pand %xmm8, %xmm2
>> ; SSE2-NEXT:    packuswb %xmm5, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm4
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = 
>> xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = 
>> xmm4[8],xmm7[8],xmm4[9],xmm7[9],xmm4[10],xmm7[10],xmm4[11],xmm7[11],xmm4[12],xmm7[12],xmm4[13],xmm7[13],xmm4[14],xmm7[14],xmm4[15],xmm7[15]
>> ; SSE2-NEXT:    psraw $8, %xmm4
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm5
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = 
>> xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = 
>> xmm5[8],xmm3[8],xmm5[9],xmm3[9],xmm5[10],xmm3[10],xmm5[11],xmm3[11],xmm5[12],xmm3[12],xmm5[13],xmm3[13],xmm5[14],xmm3[14],xmm5[15],xmm3[15]
>> ; SSE2-NEXT:    psraw $8, %xmm5
>> ; SSE2-NEXT:    pmullw %xmm4, %xmm5
>> ; SSE2-NEXT:    pand %xmm8, %xmm5
>> @@ -1006,7 +1006,7 @@ define <64 x i8> @mul_v64i8(<64 x i8> %i
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm8
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm1
>> ; SSE41-NEXT:    pmovsxbw %xmm4, %xmm9
>> -; SSE41-NEXT:    pmovsxbw %xmm1, %xmm0
>> +; SSE41-NEXT:    pmovsxbw %xmm0, %xmm0
>> ; SSE41-NEXT:    pmullw %xmm9, %xmm0
>> ; SSE41-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,255,255]
>> ; SSE41-NEXT:    pand %xmm9, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/powi.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/powi.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/powi.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/powi.ll Mon Oct  2 15:01:37 2017
>> @@ -5,7 +5,7 @@ define double @pow_wrapper(double %a) no
>> ; CHECK-LABEL: pow_wrapper:
>> ; CHECK:       # BB#0:
>> ; CHECK-NEXT:    movapd %xmm0, %xmm1
>> -; CHECK-NEXT:    mulsd %xmm1, %xmm1
>> +; CHECK-NEXT:    mulsd %xmm0, %xmm1
>> ; CHECK-NEXT:    mulsd %xmm1, %xmm0
>> ; CHECK-NEXT:    mulsd %xmm1, %xmm1
>> ; CHECK-NEXT:    mulsd %xmm1, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/pr11334.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/pr11334.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/pr11334.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/pr11334.ll Mon Oct  2 15:01:37 2017
>> @@ -25,7 +25,7 @@ define <3 x double> @v3f2d_ext_vec(<3 x
>> ; SSE-NEXT:    cvtps2pd %xmm0, %xmm0
>> ; SSE-NEXT:    movlps %xmm0, -{{[0-9]+}}(%rsp)
>> ; SSE-NEXT:    movaps %xmm2, %xmm1
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm2[1],xmm1[1]
>> ; SSE-NEXT:    fldl -{{[0-9]+}}(%rsp)
>> ; SSE-NEXT:    movaps %xmm2, %xmm0
>> ; SSE-NEXT:    retq
>>
>> Modified: llvm/trunk/test/CodeGen/X86/pr29112.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/pr29112.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/pr29112.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/pr29112.ll Mon Oct  2 15:01:37 2017
>> @@ -49,16 +49,16 @@ define <4 x float> @bar(<4 x float>* %a1
>> ; CHECK-NEXT:    vinsertps {{.*#+}} xmm2 = xmm9[0,1],xmm2[3],xmm9[3]
>> ; CHECK-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],xmm12[0]
>> ; CHECK-NEXT:    vaddps %xmm3, %xmm2, %xmm2
>> -; CHECK-NEXT:    vmovaps %xmm15, %xmm1
>> -; CHECK-NEXT:    vmovaps %xmm1, {{[0-9]+}}(%rsp) # 16-byte Spill
>> -; CHECK-NEXT:    vaddps %xmm0, %xmm1, %xmm9
>> +; CHECK-NEXT:    vmovaps %xmm15, {{[0-9]+}}(%rsp) # 16-byte Spill
>> +; CHECK-NEXT:    vaddps %xmm0, %xmm15, %xmm9
>> ; CHECK-NEXT:    vaddps %xmm14, %xmm10, %xmm0
>> -; CHECK-NEXT:    vaddps %xmm1, %xmm1, %xmm8
>> +; CHECK-NEXT:    vaddps %xmm15, %xmm15, %xmm8
>> ; CHECK-NEXT:    vaddps %xmm11, %xmm3, %xmm3
>> ; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0
>> -; CHECK-NEXT:    vaddps %xmm0, %xmm1, %xmm0
>> +; CHECK-NEXT:    vaddps %xmm0, %xmm15, %xmm0
>> ; CHECK-NEXT:    vmovaps %xmm8, {{[0-9]+}}(%rsp)
>> ; CHECK-NEXT:    vmovaps %xmm9, (%rsp)
>> +; CHECK-NEXT:    vmovaps %xmm15, %xmm1
>> ; CHECK-NEXT:    vmovaps {{[0-9]+}}(%rsp), %xmm3 # 16-byte Reload
>> ; CHECK-NEXT:    vzeroupper
>> ; CHECK-NEXT:    callq foo
>>
>> Modified: llvm/trunk/test/CodeGen/X86/psubus.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/psubus.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/psubus.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/psubus.ll Mon Oct  2 15:01:37 2017
>> @@ -712,7 +712,7 @@ define <16 x i8> @test14(<16 x i8> %x, <
>> ; SSE41-LABEL: test14:
>> ; SSE41:       # BB#0: # %vector.ph
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm5
>> -; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,2,3]
>> +; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
>> ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm8 = 
>> xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
>> ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = 
>> xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero,xmm5[2],zero,zero,zero,xmm5[3],zero,zero,zero
>> ; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[2,3,0,1]
>> @@ -1320,7 +1320,7 @@ define <32 x i16> @psubus_32i16_max(<32
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm9
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm8
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = 
>> [32768,32768,32768,32768,32768,32768,32768,32768]
>> -; SSE2-NEXT:    movdqa %xmm8, %xmm1
>> +; SSE2-NEXT:    movdqa %xmm0, %xmm1
>> ; SSE2-NEXT:    pxor %xmm3, %xmm1
>> ; SSE2-NEXT:    movdqa %xmm4, %xmm0
>> ; SSE2-NEXT:    pxor %xmm3, %xmm0
>> @@ -1368,7 +1368,7 @@ define <32 x i16> @psubus_32i16_max(<32
>> ; SSSE3-NEXT:    movdqa %xmm1, %xmm9
>> ; SSSE3-NEXT:    movdqa %xmm0, %xmm8
>> ; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = 
>> [32768,32768,32768,32768,32768,32768,32768,32768]
>> -; SSSE3-NEXT:    movdqa %xmm8, %xmm1
>> +; SSSE3-NEXT:    movdqa %xmm0, %xmm1
>> ; SSSE3-NEXT:    pxor %xmm3, %xmm1
>> ; SSSE3-NEXT:    movdqa %xmm4, %xmm0
>> ; SSSE3-NEXT:    pxor %xmm3, %xmm0
>> @@ -2042,7 +2042,7 @@ define <16 x i16> @psubus_16i32_max(<16
>> ; SSE2-NEXT:    movdqa %xmm9, %xmm11
>> ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm11 = 
>> xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7]
>> ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = 
>> xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
>> -; SSE2-NEXT:    movdqa %xmm8, %xmm10
>> +; SSE2-NEXT:    movdqa %xmm1, %xmm10
>> ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm10 = 
>> xmm10[4],xmm0[4],xmm10[5],xmm0[5],xmm10[6],xmm0[6],xmm10[7],xmm0[7]
>> ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = 
>> xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm6 = 
>> [2147483648,2147483648,2147483648,2147483648]
>> @@ -2105,7 +2105,7 @@ define <16 x i16> @psubus_16i32_max(<16
>> ; SSSE3-NEXT:    movdqa %xmm9, %xmm11
>> ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm11 = 
>> xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7]
>> ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm9 = 
>> xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
>> -; SSSE3-NEXT:    movdqa %xmm8, %xmm10
>> +; SSSE3-NEXT:    movdqa %xmm1, %xmm10
>> ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm10 = 
>> xmm10[4],xmm0[4],xmm10[5],xmm0[5],xmm10[6],xmm0[6],xmm10[7],xmm0[7]
>> ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm8 = 
>> xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
>> ; SSSE3-NEXT:    movdqa {{.*#+}} xmm6 = 
>> [2147483648,2147483648,2147483648,2147483648]
>>
>> Modified: llvm/trunk/test/CodeGen/X86/shrink-wrap-chkstk.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/shrink-wrap-chkstk.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/shrink-wrap-chkstk.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/shrink-wrap-chkstk.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -61,7 +61,7 @@ false:
>>
>> ; CHECK-LABEL: @use_eax_before_prologue at 8: # @use_eax_before_prologue
>> ; CHECK: movl %ecx, %eax
>> -; CHECK: cmpl %edx, %eax
>> +; CHECK: cmpl %edx, %ecx
>> ; CHECK: jge LBB1_2
>> ; CHECK: pushl %eax
>> ; CHECK: movl $4092, %eax
>>
>> Modified: llvm/trunk/test/CodeGen/X86/sqrt-fastmath.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/sqrt-fastmath.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/sqrt-fastmath.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/sqrt-fastmath.ll Mon Oct  2 15:01:37 2017
>> @@ -132,7 +132,7 @@ define float @f32_estimate(float %x) #1
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    rsqrtss %xmm0, %xmm1
>> ; SSE-NEXT:    movaps %xmm1, %xmm2
>> -; SSE-NEXT:    mulss %xmm2, %xmm2
>> +; SSE-NEXT:    mulss %xmm1, %xmm2
>> ; SSE-NEXT:    mulss %xmm0, %xmm2
>> ; SSE-NEXT:    addss {{.*}}(%rip), %xmm2
>> ; SSE-NEXT:    mulss {{.*}}(%rip), %xmm1
>> @@ -178,7 +178,7 @@ define <4 x float> @v4f32_estimate(<4 x
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    rsqrtps %xmm0, %xmm1
>> ; SSE-NEXT:    movaps %xmm1, %xmm2
>> -; SSE-NEXT:    mulps %xmm2, %xmm2
>> +; SSE-NEXT:    mulps %xmm1, %xmm2
>> ; SSE-NEXT:    mulps %xmm0, %xmm2
>> ; SSE-NEXT:    addps {{.*}}(%rip), %xmm2
>> ; SSE-NEXT:    mulps {{.*}}(%rip), %xmm1
>> @@ -228,7 +228,7 @@ define <8 x float> @v8f32_estimate(<8 x
>> ; SSE-NEXT:    rsqrtps %xmm0, %xmm3
>> ; SSE-NEXT:    movaps {{.*#+}} xmm4 = 
>> [-5.000000e-01,-5.000000e-01,-5.000000e-01,-5.000000e-01]
>> ; SSE-NEXT:    movaps %xmm3, %xmm2
>> -; SSE-NEXT:    mulps %xmm2, %xmm2
>> +; SSE-NEXT:    mulps %xmm3, %xmm2
>> ; SSE-NEXT:    mulps %xmm0, %xmm2
>> ; SSE-NEXT:    movaps {{.*#+}} xmm0 = 
>> [-3.000000e+00,-3.000000e+00,-3.000000e+00,-3.000000e+00]
>> ; SSE-NEXT:    addps %xmm0, %xmm2
>> @@ -236,7 +236,7 @@ define <8 x float> @v8f32_estimate(<8 x
>> ; SSE-NEXT:    mulps %xmm3, %xmm2
>> ; SSE-NEXT:    rsqrtps %xmm1, %xmm5
>> ; SSE-NEXT:    movaps %xmm5, %xmm3
>> -; SSE-NEXT:    mulps %xmm3, %xmm3
>> +; SSE-NEXT:    mulps %xmm5, %xmm3
>> ; SSE-NEXT:    mulps %xmm1, %xmm3
>> ; SSE-NEXT:    addps %xmm0, %xmm3
>> ; SSE-NEXT:    mulps %xmm4, %xmm3
>>
>> Modified: llvm/trunk/test/CodeGen/X86/sse1.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/sse1.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/sse1.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/sse1.ll Mon Oct  2 15:01:37 2017
>> @@ -16,7 +16,7 @@ define <2 x float> @test4(<2 x float> %A
>> ; X32-LABEL: test4:
>> ; X32:       # BB#0: # %entry
>> ; X32-NEXT:    movaps %xmm0, %xmm2
>> -; X32-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1,2,3]
>> +; X32-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm0[2,3]
>> ; X32-NEXT:    addss %xmm1, %xmm0
>> ; X32-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1,2,3]
>> ; X32-NEXT:    subss %xmm1, %xmm2
>> @@ -26,7 +26,7 @@ define <2 x float> @test4(<2 x float> %A
>> ; X64-LABEL: test4:
>> ; X64:       # BB#0: # %entry
>> ; X64-NEXT:    movaps %xmm0, %xmm2
>> -; X64-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1,2,3]
>> +; X64-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm0[2,3]
>> ; X64-NEXT:    addss %xmm1, %xmm0
>> ; X64-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1,2,3]
>> ; X64-NEXT:    subss %xmm1, %xmm2
>>
>> Modified: llvm/trunk/test/CodeGen/X86/sse3-avx-addsub-2.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/sse3-avx-addsub-2.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/sse3-avx-addsub-2.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/sse3-avx-addsub-2.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -406,9 +406,9 @@ define <4 x float> @test16(<4 x float> %
>> ; SSE-NEXT:    movaps %xmm0, %xmm2
>> ; SSE-NEXT:    subss %xmm0, %xmm2
>> ; SSE-NEXT:    movaps %xmm0, %xmm3
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm3 = xmm3[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm3 = xmm0[1],xmm3[1]
>> ; SSE-NEXT:    movaps %xmm1, %xmm4
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm4 = xmm4[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm4 = xmm1[1],xmm4[1]
>> ; SSE-NEXT:    subss %xmm4, %xmm3
>> ; SSE-NEXT:    movshdup {{.*#+}} xmm4 = xmm0[1,1,3,3]
>> ; SSE-NEXT:    addss %xmm0, %xmm4
>>
>> Modified: llvm/trunk/test/CodeGen/X86/statepoint-live-in.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/statepoint-live-in.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/statepoint-live-in.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/statepoint-live-in.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -126,7 +126,7 @@ define void @test6(i32 %a) gc "statepoin
>> ; CHECK-NEXT:  Lcfi11:
>> ; CHECK-NEXT:    .cfi_offset %rbx, -16
>> ; CHECK-NEXT:    movl %edi, %ebx
>> -; CHECK-NEXT:    movl %ebx, {{[0-9]+}}(%rsp)
>> +; CHECK-NEXT:    movl %edi, {{[0-9]+}}(%rsp)
>> ; CHECK-NEXT:    callq _baz
>> ; CHECK-NEXT:  Ltmp6:
>> ; CHECK-NEXT:    callq _bar
>> @@ -153,13 +153,13 @@ entry:
>> ; CHECK:      .byte1
>> ; CHECK-NEXT:   .byte   0
>> ; CHECK-NEXT: .short 4
>> -; CHECK-NEXT: .short6
>> +; CHECK-NEXT: .short5
>> ; CHECK-NEXT:   .short  0
>> ; CHECK-NEXT: .long0
>> ; CHECK:      .byte1
>> ; CHECK-NEXT:   .byte   0
>> ; CHECK-NEXT: .short 4
>> -; CHECK-NEXT: .short3
>> +; CHECK-NEXT: .short4
>> ; CHECK-NEXT:   .short  0
>> ; CHECK-NEXT: .long0
>> ; CHECK: Ltmp2-_test2
>>
>> Modified: llvm/trunk/test/CodeGen/X86/statepoint-stack-usage.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/statepoint-stack-usage.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/statepoint-stack-usage.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/statepoint-stack-usage.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -61,9 +61,9 @@ define i32 @back_to_back_deopt(i32 %a, i
>>   gc "statepoint-example" {
>> ; CHECK-LABEL: back_to_back_deopt
>> ; The exact stores don't matter, but there need to be three stack 
>> slots created
>> -; CHECK-DAG: movl%ebx, 12(%rsp)
>> -; CHECK-DAG: movl%ebp, 8(%rsp)
>> -; CHECK-DAG: movl%r14d, 4(%rsp)
>> +; CHECK-DAG: movl%edi, 12(%rsp)
>> +; CHECK-DAG: movl%esi, 8(%rsp)
>> +; CHECK-DAG: movl%edx, 4(%rsp)
>> ; CHECK: callq
>> ; CHECK-DAG: movl%ebx, 12(%rsp)
>> ; CHECK-DAG: movl%ebp, 8(%rsp)
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vec_fp_to_int.ll Mon Oct  2 15:01:37 2017
>> @@ -1018,12 +1018,12 @@ define <4 x i64> @fptosi_4f32_to_4i64(<8
>> ; SSE-NEXT:    cvttss2si %xmm0, %rax
>> ; SSE-NEXT:    movq %rax, %xmm2
>> ; SSE-NEXT:    movaps %xmm0, %xmm1
>> -; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[2,3]
>> ; SSE-NEXT:    cvttss2si %xmm1, %rax
>> ; SSE-NEXT:    movq %rax, %xmm1
>> ; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
>> ; SSE-NEXT:    movaps %xmm0, %xmm1
>> -; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
>> ; SSE-NEXT:    cvttss2si %xmm1, %rax
>> ; SSE-NEXT:    movq %rax, %xmm3
>> ; SSE-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
>> @@ -1126,12 +1126,12 @@ define <4 x i64> @fptosi_8f32_to_4i64(<8
>> ; SSE-NEXT:    cvttss2si %xmm0, %rax
>> ; SSE-NEXT:    movq %rax, %xmm2
>> ; SSE-NEXT:    movaps %xmm0, %xmm1
>> -; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[2,3]
>> ; SSE-NEXT:    cvttss2si %xmm1, %rax
>> ; SSE-NEXT:    movq %rax, %xmm1
>> ; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
>> ; SSE-NEXT:    movaps %xmm0, %xmm1
>> -; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
>> ; SSE-NEXT:    cvttss2si %xmm1, %rax
>> ; SSE-NEXT:    movq %rax, %xmm3
>> ; SSE-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
>> @@ -1316,11 +1316,11 @@ define <4 x i32> @fptoui_4f32_to_4i32(<4
>> ; SSE-LABEL: fptoui_4f32_to_4i32:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movaps %xmm0, %xmm1
>> -; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
>> ; SSE-NEXT:    cvttss2si %xmm1, %rax
>> ; SSE-NEXT:    movd %eax, %xmm1
>> ; SSE-NEXT:    movaps %xmm0, %xmm2
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm2[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm0[1],xmm2[1]
>> ; SSE-NEXT:    cvttss2si %xmm2, %rax
>> ; SSE-NEXT:    movd %eax, %xmm2
>> ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
>> @@ -1560,7 +1560,7 @@ define <8 x i32> @fptoui_8f32_to_8i32(<8
>> ; SSE-NEXT:    cvttss2si %xmm0, %rax
>> ; SSE-NEXT:    movd %eax, %xmm0
>> ; SSE-NEXT:    movaps %xmm2, %xmm3
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm3 = xmm3[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm3 = xmm2[1],xmm3[1]
>> ; SSE-NEXT:    cvttss2si %xmm3, %rax
>> ; SSE-NEXT:    movd %eax, %xmm3
>> ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
>> @@ -1572,11 +1572,11 @@ define <8 x i32> @fptoui_8f32_to_8i32(<8
>> ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
>> ; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
>> ; SSE-NEXT:    movaps %xmm1, %xmm2
>> -; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm1[2,3]
>> ; SSE-NEXT:    cvttss2si %xmm2, %rax
>> ; SSE-NEXT:    movd %eax, %xmm2
>> ; SSE-NEXT:    movaps %xmm1, %xmm3
>> -; SSE-NEXT:    movhlps {{.*#+}} xmm3 = xmm3[1,1]
>> +; SSE-NEXT:    movhlps {{.*#+}} xmm3 = xmm1[1],xmm3[1]
>> ; SSE-NEXT:    cvttss2si %xmm3, %rax
>> ; SSE-NEXT:    movd %eax, %xmm3
>> ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
>> @@ -1687,7 +1687,7 @@ define <4 x i64> @fptoui_4f32_to_4i64(<8
>> ; SSE-NEXT:    cmovaeq %rcx, %rdx
>> ; SSE-NEXT:    movq %rdx, %xmm2
>> ; SSE-NEXT:    movaps %xmm0, %xmm3
>> -; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[2,3]
>> ; SSE-NEXT:    movaps %xmm3, %xmm4
>> ; SSE-NEXT:    subss %xmm1, %xmm4
>> ; SSE-NEXT:    cvttss2si %xmm4, %rcx
>> @@ -1698,7 +1698,7 @@ define <4 x i64> @fptoui_4f32_to_4i64(<8
>> ; SSE-NEXT:    movq %rdx, %xmm3
>> ; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
>> ; SSE-NEXT:    movaps %xmm0, %xmm3
>> -; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
>> ; SSE-NEXT:    movaps %xmm3, %xmm4
>> ; SSE-NEXT:    subss %xmm1, %xmm4
>> ; SSE-NEXT:    cvttss2si %xmm4, %rcx
>> @@ -1865,7 +1865,7 @@ define <4 x i64> @fptoui_8f32_to_4i64(<8
>> ; SSE-NEXT:    cmovaeq %rcx, %rdx
>> ; SSE-NEXT:    movq %rdx, %xmm2
>> ; SSE-NEXT:    movaps %xmm0, %xmm3
>> -; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[2,3]
>> ; SSE-NEXT:    movaps %xmm3, %xmm4
>> ; SSE-NEXT:    subss %xmm1, %xmm4
>> ; SSE-NEXT:    cvttss2si %xmm4, %rcx
>> @@ -1876,7 +1876,7 @@ define <4 x i64> @fptoui_8f32_to_4i64(<8
>> ; SSE-NEXT:    movq %rdx, %xmm3
>> ; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
>> ; SSE-NEXT:    movaps %xmm0, %xmm3
>> -; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1,2,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
>> ; SSE-NEXT:    movaps %xmm3, %xmm4
>> ; SSE-NEXT:    subss %xmm1, %xmm4
>> ; SSE-NEXT:    cvttss2si %xmm4, %rcx
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vec_int_to_fp.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vec_int_to_fp.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vec_int_to_fp.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vec_int_to_fp.ll Mon Oct  2 15:01:37 2017
>> @@ -1591,7 +1591,7 @@ define <4 x float> @uitofp_2i64_to_4f32(
>> ; SSE-LABEL: uitofp_2i64_to_4f32:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE-NEXT:    movq %xmm1, %rax
>> +; SSE-NEXT:    movq %xmm0, %rax
>> ; SSE-NEXT:    testq %rax, %rax
>> ; SSE-NEXT:    js .LBB39_1
>> ; SSE-NEXT:  # BB#2:
>> @@ -1819,7 +1819,7 @@ define <4 x float> @uitofp_4i64_to_4f32_
>> ; SSE-LABEL: uitofp_4i64_to_4f32_undef:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE-NEXT:    movq %xmm1, %rax
>> +; SSE-NEXT:    movq %xmm0, %rax
>> ; SSE-NEXT:    testq %rax, %rax
>> ; SSE-NEXT:    js .LBB41_1
>> ; SSE-NEXT:  # BB#2:
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vec_minmax_sint.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vec_minmax_sint.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vec_minmax_sint.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vec_minmax_sint.ll Mon Oct  2 15:01:37 
>> 2017
>> @@ -437,7 +437,7 @@ define <2 x i64> @max_ge_v2i64(<2 x i64>
>> ; SSE42:       # BB#0:
>> ; SSE42-NEXT:    movdqa %xmm0, %xmm2
>> ; SSE42-NEXT:    movdqa %xmm1, %xmm3
>> -; SSE42-NEXT:    pcmpgtq %xmm2, %xmm3
>> +; SSE42-NEXT:    pcmpgtq %xmm0, %xmm3
>> ; SSE42-NEXT:    pcmpeqd %xmm0, %xmm0
>> ; SSE42-NEXT:    pxor %xmm3, %xmm0
>> ; SSE42-NEXT:    blendvpd %xmm0, %xmm2, %xmm1
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vec_shift4.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vec_shift4.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vec_shift4.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vec_shift4.ll Mon Oct  2 15:01:37 2017
>> @@ -35,7 +35,7 @@ define <2 x i64> @shl2(<16 x i8> %r, <16
>> ; X32:       # BB#0: # %entry
>> ; X32-NEXT:    movdqa %xmm0, %xmm2
>> ; X32-NEXT:    psllw $5, %xmm1
>> -; X32-NEXT:    movdqa %xmm2, %xmm3
>> +; X32-NEXT:    movdqa %xmm0, %xmm3
>> ; X32-NEXT:    psllw $4, %xmm3
>> ; X32-NEXT:    pand {{\.LCPI.*}}, %xmm3
>> ; X32-NEXT:    movdqa %xmm1, %xmm0
>> @@ -47,7 +47,7 @@ define <2 x i64> @shl2(<16 x i8> %r, <16
>> ; X32-NEXT:    movdqa %xmm1, %xmm0
>> ; X32-NEXT:    pblendvb %xmm0, %xmm3, %xmm2
>> ; X32-NEXT:    movdqa %xmm2, %xmm3
>> -; X32-NEXT:    paddb %xmm3, %xmm3
>> +; X32-NEXT:    paddb %xmm2, %xmm3
>> ; X32-NEXT:    paddb %xmm1, %xmm1
>> ; X32-NEXT:    movdqa %xmm1, %xmm0
>> ; X32-NEXT:    pblendvb %xmm0, %xmm3, %xmm2
>> @@ -58,7 +58,7 @@ define <2 x i64> @shl2(<16 x i8> %r, <16
>> ; X64:       # BB#0: # %entry
>> ; X64-NEXT:    movdqa %xmm0, %xmm2
>> ; X64-NEXT:    psllw $5, %xmm1
>> -; X64-NEXT:    movdqa %xmm2, %xmm3
>> +; X64-NEXT:    movdqa %xmm0, %xmm3
>> ; X64-NEXT:    psllw $4, %xmm3
>> ; X64-NEXT:    pand {{.*}}(%rip), %xmm3
>> ; X64-NEXT:    movdqa %xmm1, %xmm0
>> @@ -70,7 +70,7 @@ define <2 x i64> @shl2(<16 x i8> %r, <16
>> ; X64-NEXT:    movdqa %xmm1, %xmm0
>> ; X64-NEXT:    pblendvb %xmm0, %xmm3, %xmm2
>> ; X64-NEXT:    movdqa %xmm2, %xmm3
>> -; X64-NEXT:    paddb %xmm3, %xmm3
>> +; X64-NEXT:    paddb %xmm2, %xmm3
>> ; X64-NEXT:    paddb %xmm1, %xmm1
>> ; X64-NEXT:    movdqa %xmm1, %xmm0
>> ; X64-NEXT:    pblendvb %xmm0, %xmm3, %xmm2
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-blend.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-blend.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-blend.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-blend.ll Mon Oct  2 15:01:37 2017
>> @@ -986,7 +986,7 @@ define <4 x i32> @blend_neg_logic_v4i32_
>> ; SSE41:       # BB#0: # %entry
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm2
>> ; SSE41-NEXT:    pxor %xmm3, %xmm3
>> -; SSE41-NEXT:    psubd %xmm2, %xmm3
>> +; SSE41-NEXT:    psubd %xmm0, %xmm3
>> ; SSE41-NEXT:    movaps %xmm1, %xmm0
>> ; SSE41-NEXT:    blendvps %xmm0, %xmm2, %xmm3
>> ; SSE41-NEXT:    movaps %xmm3, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-idiv-sdiv-128.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-idiv-sdiv-128.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-idiv-sdiv-128.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-idiv-sdiv-128.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -177,13 +177,13 @@ define <16 x i8> @test_div7_16i8(<16 x i
>> ; SSE2-LABEL: test_div7_16i8:
>> ; SSE2:       # BB#0:
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm2
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
>> ; SSE2-NEXT:    psraw $8, %xmm2
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = 
>> [65427,65427,65427,65427,65427,65427,65427,65427]
>> ; SSE2-NEXT:    pmullw %xmm3, %xmm2
>> ; SSE2-NEXT:    psrlw $8, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = 
>> xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
>> +; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = 
>> xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
>> ; SSE2-NEXT:    psraw $8, %xmm1
>> ; SSE2-NEXT:    pmullw %xmm3, %xmm1
>> ; SSE2-NEXT:    psrlw $8, %xmm1
>> @@ -501,13 +501,13 @@ define <16 x i8> @test_rem7_16i8(<16 x i
>> ; SSE2-LABEL: test_rem7_16i8:
>> ; SSE2:       # BB#0:
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm2
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
>> ; SSE2-NEXT:    psraw $8, %xmm2
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = 
>> [65427,65427,65427,65427,65427,65427,65427,65427]
>> ; SSE2-NEXT:    pmullw %xmm3, %xmm2
>> ; SSE2-NEXT:    psrlw $8, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = 
>> xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
>> +; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = 
>> xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
>> ; SSE2-NEXT:    psraw $8, %xmm1
>> ; SSE2-NEXT:    pmullw %xmm3, %xmm1
>> ; SSE2-NEXT:    psrlw $8, %xmm1
>> @@ -523,7 +523,7 @@ define <16 x i8> @test_rem7_16i8(<16 x i
>> ; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
>> ; SSE2-NEXT:    paddb %xmm2, %xmm1
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
>> ; SSE2-NEXT:    psraw $8, %xmm2
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7]
>> ; SSE2-NEXT:    pmullw %xmm3, %xmm2
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-idiv-udiv-128.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-idiv-udiv-128.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-idiv-udiv-128.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-idiv-udiv-128.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -497,7 +497,7 @@ define <16 x i8> @test_rem7_16i8(<16 x i
>> ; SSE2-NEXT:    psrlw $2, %xmm1
>> ; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> -; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
>> +; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = 
>> xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
>> ; SSE2-NEXT:    psraw $8, %xmm2
>> ; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [7,7,7,7,7,7,7,7]
>> ; SSE2-NEXT:    pmullw %xmm3, %xmm2
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-mul.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-mul.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-mul.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-mul.ll Mon Oct  2 15:01:37 2017
>> @@ -178,7 +178,7 @@ define <16 x i8> @mul_v16i8_1_2_4_8_1_2_
>> ; X86-LABEL: mul_v16i8_1_2_4_8_1_2_4_8_1_2_4_8_1_2_4_8:
>> ; X86:       # BB#0:
>> ; X86-NEXT:    movdqa %xmm0, %xmm1
>> -; X86-NEXT:    movdqa %xmm1, %xmm2
>> +; X86-NEXT:    movdqa %xmm0, %xmm2
>> ; X86-NEXT:    psllw $4, %xmm2
>> ; X86-NEXT:    pand {{\.LCPI.*}}, %xmm2
>> ; X86-NEXT:    movdqa {{.*#+}} xmm0 = 
>> [8192,24640,8192,24640,8192,24640,8192,24640]
>> @@ -189,7 +189,7 @@ define <16 x i8> @mul_v16i8_1_2_4_8_1_2_
>> ; X86-NEXT:    paddb %xmm0, %xmm0
>> ; X86-NEXT:    pblendvb %xmm0, %xmm2, %xmm1
>> ; X86-NEXT:    movdqa %xmm1, %xmm2
>> -; X86-NEXT:    paddb %xmm2, %xmm2
>> +; X86-NEXT:    paddb %xmm1, %xmm2
>> ; X86-NEXT:    paddb %xmm0, %xmm0
>> ; X86-NEXT:    pblendvb %xmm0, %xmm2, %xmm1
>> ; X86-NEXT:    movdqa %xmm1, %xmm0
>> @@ -198,7 +198,7 @@ define <16 x i8> @mul_v16i8_1_2_4_8_1_2_
>> ; X64-LABEL: mul_v16i8_1_2_4_8_1_2_4_8_1_2_4_8_1_2_4_8:
>> ; X64:       # BB#0:
>> ; X64-NEXT:    movdqa %xmm0, %xmm1
>> -; X64-NEXT:    movdqa %xmm1, %xmm2
>> +; X64-NEXT:    movdqa %xmm0, %xmm2
>> ; X64-NEXT:    psllw $4, %xmm2
>> ; X64-NEXT:    pand {{.*}}(%rip), %xmm2
>> ; X64-NEXT:    movdqa {{.*#+}} xmm0 = 
>> [8192,24640,8192,24640,8192,24640,8192,24640]
>> @@ -209,7 +209,7 @@ define <16 x i8> @mul_v16i8_1_2_4_8_1_2_
>> ; X64-NEXT:    paddb %xmm0, %xmm0
>> ; X64-NEXT:    pblendvb %xmm0, %xmm2, %xmm1
>> ; X64-NEXT:    movdqa %xmm1, %xmm2
>> -; X64-NEXT:    paddb %xmm2, %xmm2
>> +; X64-NEXT:    paddb %xmm1, %xmm2
>> ; X64-NEXT:    paddb %xmm0, %xmm0
>> ; X64-NEXT:    pblendvb %xmm0, %xmm2, %xmm1
>> ; X64-NEXT:    movdqa %xmm1, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-rotate-128.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-rotate-128.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-rotate-128.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-rotate-128.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -361,7 +361,7 @@ define <8 x i16> @var_rotate_v8i16(<8 x
>> ; SSE41-NEXT:    psllw $4, %xmm1
>> ; SSE41-NEXT:    por %xmm0, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm4
>> -; SSE41-NEXT:    paddw %xmm4, %xmm4
>> +; SSE41-NEXT:    paddw %xmm1, %xmm4
>> ; SSE41-NEXT:    movdqa %xmm3, %xmm6
>> ; SSE41-NEXT:    psllw $8, %xmm6
>> ; SSE41-NEXT:    movdqa %xmm3, %xmm5
>> @@ -386,7 +386,7 @@ define <8 x i16> @var_rotate_v8i16(<8 x
>> ; SSE41-NEXT:    psllw $4, %xmm2
>> ; SSE41-NEXT:    por %xmm0, %xmm2
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm1
>> -; SSE41-NEXT:    paddw %xmm1, %xmm1
>> +; SSE41-NEXT:    paddw %xmm2, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm3, %xmm4
>> ; SSE41-NEXT:    psrlw $8, %xmm4
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm0
>> @@ -631,10 +631,10 @@ define <16 x i8> @var_rotate_v16i8(<16 x
>> ; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
>> ; SSE41-NEXT:    psubb %xmm3, %xmm2
>> ; SSE41-NEXT:    psllw $5, %xmm3
>> -; SSE41-NEXT:    movdqa %xmm1, %xmm5
>> +; SSE41-NEXT:    movdqa %xmm0, %xmm5
>> ; SSE41-NEXT:    psllw $4, %xmm5
>> ; SSE41-NEXT:    pand {{.*}}(%rip), %xmm5
>> -; SSE41-NEXT:    movdqa %xmm1, %xmm4
>> +; SSE41-NEXT:    movdqa %xmm0, %xmm4
>> ; SSE41-NEXT:    movdqa %xmm3, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm5, %xmm4
>> ; SSE41-NEXT:    movdqa %xmm4, %xmm5
>> @@ -644,13 +644,13 @@ define <16 x i8> @var_rotate_v16i8(<16 x
>> ; SSE41-NEXT:    movdqa %xmm3, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm5, %xmm4
>> ; SSE41-NEXT:    movdqa %xmm4, %xmm5
>> -; SSE41-NEXT:    paddb %xmm5, %xmm5
>> +; SSE41-NEXT:    paddb %xmm4, %xmm5
>> ; SSE41-NEXT:    paddb %xmm3, %xmm3
>> ; SSE41-NEXT:    movdqa %xmm3, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm5, %xmm4
>> ; SSE41-NEXT:    psllw $5, %xmm2
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm3
>> -; SSE41-NEXT:    paddb %xmm3, %xmm3
>> +; SSE41-NEXT:    paddb %xmm2, %xmm3
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm5
>> ; SSE41-NEXT:    psrlw $4, %xmm5
>> ; SSE41-NEXT:    pand {{.*}}(%rip), %xmm5
>> @@ -1191,7 +1191,7 @@ define <16 x i8> @constant_rotate_v16i8(
>> ; SSE41-LABEL: constant_rotate_v16i8:
>> ; SSE41:       # BB#0:
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE41-NEXT:    movdqa %xmm1, %xmm3
>> +; SSE41-NEXT:    movdqa %xmm0, %xmm3
>> ; SSE41-NEXT:    psllw $4, %xmm3
>> ; SSE41-NEXT:    pand {{.*}}(%rip), %xmm3
>> ; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = 
>> [8192,24640,41088,57536,57600,41152,24704,8256]
>> @@ -1203,7 +1203,7 @@ define <16 x i8> @constant_rotate_v16i8(
>> ; SSE41-NEXT:    paddb %xmm0, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm3, %xmm2
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm3
>> -; SSE41-NEXT:    paddb %xmm3, %xmm3
>> +; SSE41-NEXT:    paddb %xmm2, %xmm3
>> ; SSE41-NEXT:    paddb %xmm0, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm3, %xmm2
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm3
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-sext.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-sext.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-sext.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-sext.ll Mon Oct  2 15:01:37 2017
>> @@ -243,7 +243,7 @@ define <8 x i32> @sext_16i8_to_8i32(<16
>> ; SSSE3-LABEL: sext_16i8_to_8i32:
>> ; SSSE3:       # BB#0: # %entry
>> ; SSSE3-NEXT:    movdqa %xmm0, %xmm1
>> -; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = 
>> xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
>> +; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = 
>> xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
>> ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
>> ; SSSE3-NEXT:    psrad $24, %xmm0
>> ; SSSE3-NEXT:    pshufb {{.*#+}} xmm1 = 
>> xmm1[u,u,u,4,u,u,u,5,u,u,u,6,u,u,u,7]
>> @@ -312,7 +312,7 @@ define <16 x i32> @sext_16i8_to_16i32(<1
>> ; SSSE3-LABEL: sext_16i8_to_16i32:
>> ; SSSE3:       # BB#0: # %entry
>> ; SSSE3-NEXT:    movdqa %xmm0, %xmm3
>> -; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = 
>> xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
>> +; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = 
>> xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
>> ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
>> ; SSSE3-NEXT:    psrad $24, %xmm0
>> ; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm1 = 
>> xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
>> @@ -443,7 +443,7 @@ define <4 x i64> @sext_16i8_to_4i64(<16
>> ; SSSE3-LABEL: sext_16i8_to_4i64:
>> ; SSSE3:       # BB#0: # %entry
>> ; SSSE3-NEXT:    movdqa %xmm0, %xmm1
>> -; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = 
>> xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
>> +; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = 
>> xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
>> ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
>> ; SSSE3-NEXT:    movdqa %xmm0, %xmm2
>> ; SSSE3-NEXT:    psrad $31, %xmm2
>> @@ -499,7 +499,7 @@ define <8 x i64> @sext_16i8_to_8i64(<16
>> ; SSE2-LABEL: sext_16i8_to_8i64:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = 
>> xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
>> +; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = 
>> xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
>> ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm2
>> ; SSE2-NEXT:    psrad $31, %xmm2
>> @@ -1112,7 +1112,7 @@ define <8 x i64> @sext_8i32_to_8i64(<8 x
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm3
>> ; SSE2-NEXT:    psrad $31, %xmm3
>> -; SSE2-NEXT:    movdqa %xmm2, %xmm4
>> +; SSE2-NEXT:    movdqa %xmm1, %xmm4
>> ; SSE2-NEXT:    psrad $31, %xmm4
>> ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
>> ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
>> @@ -1131,7 +1131,7 @@ define <8 x i64> @sext_8i32_to_8i64(<8 x
>> ; SSSE3-NEXT:    movdqa %xmm1, %xmm2
>> ; SSSE3-NEXT:    movdqa %xmm0, %xmm3
>> ; SSSE3-NEXT:    psrad $31, %xmm3
>> -; SSSE3-NEXT:    movdqa %xmm2, %xmm4
>> +; SSSE3-NEXT:    movdqa %xmm1, %xmm4
>> ; SSSE3-NEXT:    psrad $31, %xmm4
>> ; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
>> ; SSSE3-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-shift-ashr-128.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -275,7 +275,7 @@ define <8 x i16> @var_shift_v8i16(<8 x i
>> ; SSE41-NEXT:    psllw $4, %xmm1
>> ; SSE41-NEXT:    por %xmm0, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm3
>> -; SSE41-NEXT:    paddw %xmm3, %xmm3
>> +; SSE41-NEXT:    paddw %xmm1, %xmm3
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm4
>> ; SSE41-NEXT:    psraw $8, %xmm4
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-shift-lshr-128.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -245,7 +245,7 @@ define <8 x i16> @var_shift_v8i16(<8 x i
>> ; SSE41-NEXT:    psllw $4, %xmm1
>> ; SSE41-NEXT:    por %xmm0, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm3
>> -; SSE41-NEXT:    paddw %xmm3, %xmm3
>> +; SSE41-NEXT:    paddw %xmm1, %xmm3
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm4
>> ; SSE41-NEXT:    psrlw $8, %xmm4
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm0
>> @@ -410,7 +410,7 @@ define <16 x i8> @var_shift_v16i8(<16 x
>> ; SSE41:       # BB#0:
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm2
>> ; SSE41-NEXT:    psllw $5, %xmm1
>> -; SSE41-NEXT:    movdqa %xmm2, %xmm3
>> +; SSE41-NEXT:    movdqa %xmm0, %xmm3
>> ; SSE41-NEXT:    psrlw $4, %xmm3
>> ; SSE41-NEXT:    pand {{.*}}(%rip), %xmm3
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm0
>> @@ -684,7 +684,7 @@ define <16 x i8> @splatvar_shift_v16i8(<
>> ; SSE41-NEXT:    pshufb %xmm0, %xmm1
>> ; SSE41-NEXT:    psllw $5, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm3
>> -; SSE41-NEXT:    paddb %xmm3, %xmm3
>> +; SSE41-NEXT:    paddb %xmm1, %xmm3
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm4
>> ; SSE41-NEXT:    psrlw $4, %xmm4
>> ; SSE41-NEXT:    pand {{.*}}(%rip), %xmm4
>> @@ -1111,7 +1111,7 @@ define <16 x i8> @constant_shift_v16i8(<
>> ; SSE41-LABEL: constant_shift_v16i8:
>> ; SSE41:       # BB#0:
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE41-NEXT:    movdqa %xmm1, %xmm2
>> +; SSE41-NEXT:    movdqa %xmm0, %xmm2
>> ; SSE41-NEXT:    psrlw $4, %xmm2
>> ; SSE41-NEXT:    pand {{.*}}(%rip), %xmm2
>> ; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = 
>> [8192,24640,41088,57536,49376,32928,16480,32]
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-shift-shl-128.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shift-shl-128.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-shift-shl-128.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-shift-shl-128.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -202,7 +202,7 @@ define <8 x i16> @var_shift_v8i16(<8 x i
>> ; SSE41-NEXT:    psllw $4, %xmm1
>> ; SSE41-NEXT:    por %xmm0, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm3
>> -; SSE41-NEXT:    paddw %xmm3, %xmm3
>> +; SSE41-NEXT:    paddw %xmm1, %xmm3
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm4
>> ; SSE41-NEXT:    psllw $8, %xmm4
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm0
>> @@ -364,7 +364,7 @@ define <16 x i8> @var_shift_v16i8(<16 x
>> ; SSE41:       # BB#0:
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm2
>> ; SSE41-NEXT:    psllw $5, %xmm1
>> -; SSE41-NEXT:    movdqa %xmm2, %xmm3
>> +; SSE41-NEXT:    movdqa %xmm0, %xmm3
>> ; SSE41-NEXT:    psllw $4, %xmm3
>> ; SSE41-NEXT:    pand {{.*}}(%rip), %xmm3
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm0
>> @@ -376,7 +376,7 @@ define <16 x i8> @var_shift_v16i8(<16 x
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm3, %xmm2
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm3
>> -; SSE41-NEXT:    paddb %xmm3, %xmm3
>> +; SSE41-NEXT:    paddb %xmm2, %xmm3
>> ; SSE41-NEXT:    paddb %xmm1, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm3, %xmm2
>> @@ -632,7 +632,7 @@ define <16 x i8> @splatvar_shift_v16i8(<
>> ; SSE41-NEXT:    pshufb %xmm0, %xmm1
>> ; SSE41-NEXT:    psllw $5, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm3
>> -; SSE41-NEXT:    paddb %xmm3, %xmm3
>> +; SSE41-NEXT:    paddb %xmm1, %xmm3
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm4
>> ; SSE41-NEXT:    psllw $4, %xmm4
>> ; SSE41-NEXT:    pand {{.*}}(%rip), %xmm4
>> @@ -644,7 +644,7 @@ define <16 x i8> @splatvar_shift_v16i8(<
>> ; SSE41-NEXT:    movdqa %xmm3, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm1, %xmm2
>> ; SSE41-NEXT:    movdqa %xmm2, %xmm1
>> -; SSE41-NEXT:    paddb %xmm1, %xmm1
>> +; SSE41-NEXT:    paddb %xmm2, %xmm1
>> ; SSE41-NEXT:    paddb %xmm3, %xmm3
>> ; SSE41-NEXT:    movdqa %xmm3, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm1, %xmm2
>> @@ -965,7 +965,7 @@ define <16 x i8> @constant_shift_v16i8(<
>> ; SSE41-LABEL: constant_shift_v16i8:
>> ; SSE41:       # BB#0:
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm1
>> -; SSE41-NEXT:    movdqa %xmm1, %xmm2
>> +; SSE41-NEXT:    movdqa %xmm0, %xmm2
>> ; SSE41-NEXT:    psllw $4, %xmm2
>> ; SSE41-NEXT:    pand {{.*}}(%rip), %xmm2
>> ; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = 
>> [8192,24640,41088,57536,49376,32928,16480,32]
>> @@ -976,7 +976,7 @@ define <16 x i8> @constant_shift_v16i8(<
>> ; SSE41-NEXT:    paddb %xmm0, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm2, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm2
>> -; SSE41-NEXT:    paddb %xmm2, %xmm2
>> +; SSE41-NEXT:    paddb %xmm1, %xmm2
>> ; SSE41-NEXT:    paddb %xmm0, %xmm0
>> ; SSE41-NEXT:    pblendvb %xmm0, %xmm2, %xmm1
>> ; SSE41-NEXT:    movdqa %xmm1, %xmm0
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-shuffle-combining.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shuffle-combining.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-shuffle-combining.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-shuffle-combining.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -2783,7 +2783,7 @@ define <4 x float> @PR22377(<4 x float>
>> ; SSE-LABEL: PR22377:
>> ; SSE:       # BB#0: # %entry
>> ; SSE-NEXT:    movaps %xmm0, %xmm1
>> -; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,3,1,3]
>> +; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,3],xmm0[1,3]
>> ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2,0,2]
>> ; SSE-NEXT:    addps %xmm0, %xmm1
>> ; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-trunc-math.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-trunc-math.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-trunc-math.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-trunc-math.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -4964,7 +4964,7 @@ define <4 x i32> @mul_add_const_v4i64_v4
>> ; SSE-LABEL: mul_add_const_v4i64_v4i32:
>> ; SSE:       # BB#0:
>> ; SSE-NEXT:    movdqa %xmm0, %xmm2
>> -; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,1,3]
>> +; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
>> ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,3]
>> ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,1,1,3]
>> ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,3]
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vector-zext.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-zext.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vector-zext.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vector-zext.ll Mon Oct  2 15:01:37 2017
>> @@ -246,7 +246,7 @@ define <16 x i32> @zext_16i8_to_16i32(<1
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm3
>> ; SSE2-NEXT:    pxor %xmm4, %xmm4
>> -; SSE2-NEXT:    movdqa %xmm3, %xmm1
>> +; SSE2-NEXT:    movdqa %xmm0, %xmm1
>> ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = 
>> xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm0
>> ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = 
>> xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
>> @@ -261,7 +261,7 @@ define <16 x i32> @zext_16i8_to_16i32(<1
>> ; SSSE3:       # BB#0: # %entry
>> ; SSSE3-NEXT:    movdqa %xmm0, %xmm3
>> ; SSSE3-NEXT:    pxor %xmm4, %xmm4
>> -; SSSE3-NEXT:    movdqa %xmm3, %xmm1
>> +; SSSE3-NEXT:    movdqa %xmm0, %xmm1
>> ; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = 
>> xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
>> ; SSSE3-NEXT:    movdqa %xmm1, %xmm0
>> ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = 
>> xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
>> @@ -399,7 +399,7 @@ define <8 x i64> @zext_16i8_to_8i64(<16
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm1
>> ; SSE2-NEXT:    pxor %xmm4, %xmm4
>> -; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,2,3]
>> +; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,2,3]
>> ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = 
>> xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
>> ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = 
>> xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm0
>> @@ -700,7 +700,7 @@ define <8 x i64> @zext_8i16_to_8i64(<8 x
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm3
>> ; SSE2-NEXT:    pxor %xmm4, %xmm4
>> -; SSE2-NEXT:    movdqa %xmm3, %xmm1
>> +; SSE2-NEXT:    movdqa %xmm0, %xmm1
>> ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = 
>> xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm0
>> ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
>> @@ -715,7 +715,7 @@ define <8 x i64> @zext_8i16_to_8i64(<8 x
>> ; SSSE3:       # BB#0: # %entry
>> ; SSSE3-NEXT:    movdqa %xmm0, %xmm3
>> ; SSSE3-NEXT:    pxor %xmm4, %xmm4
>> -; SSSE3-NEXT:    movdqa %xmm3, %xmm1
>> +; SSSE3-NEXT:    movdqa %xmm0, %xmm1
>> ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm1 = 
>> xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
>> ; SSSE3-NEXT:    movdqa %xmm1, %xmm0
>> ; SSSE3-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
>> @@ -1582,7 +1582,7 @@ define <8 x i32> @shuf_zext_8i16_to_8i32
>> ; SSE41:       # BB#0: # %entry
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm1
>> ; SSE41-NEXT:    pxor %xmm2, %xmm2
>> -; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = 
>> xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
>> +; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = 
>> xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
>> ; SSE41-NEXT:    punpckhwd {{.*#+}} xmm1 = 
>> xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
>> ; SSE41-NEXT:    retq
>> ;
>> @@ -1630,7 +1630,7 @@ define <4 x i64> @shuf_zext_4i32_to_4i64
>> ; SSE41:       # BB#0: # %entry
>> ; SSE41-NEXT:    movdqa %xmm0, %xmm1
>> ; SSE41-NEXT:    pxor %xmm2, %xmm2
>> -; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero
>> +; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
>> ; SSE41-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
>> ; SSE41-NEXT:    retq
>> ;
>>
>> Modified: llvm/trunk/test/CodeGen/X86/vselect-minmax.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vselect-minmax.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/vselect-minmax.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/vselect-minmax.ll Mon Oct  2 15:01:37 2017
>> @@ -3344,12 +3344,12 @@ define <64 x i8> @test98(<64 x i8> %a, <
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm8
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm9
>> -; SSE2-NEXT:    movdqa %xmm8, %xmm12
>> +; SSE2-NEXT:    movdqa %xmm3, %xmm12
>> ; SSE2-NEXT:    pcmpgtb %xmm7, %xmm12
>> ; SSE2-NEXT:    pcmpeqd %xmm13, %xmm13
>> ; SSE2-NEXT:    movdqa %xmm12, %xmm3
>> ; SSE2-NEXT:    pxor %xmm13, %xmm3
>> -; SSE2-NEXT:    movdqa %xmm9, %xmm14
>> +; SSE2-NEXT:    movdqa %xmm2, %xmm14
>> ; SSE2-NEXT:    pcmpgtb %xmm6, %xmm14
>> ; SSE2-NEXT:    movdqa %xmm14, %xmm2
>> ; SSE2-NEXT:    pxor %xmm13, %xmm2
>> @@ -3487,12 +3487,12 @@ define <64 x i8> @test100(<64 x i8> %a,
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm9
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm10
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm12
>> -; SSE2-NEXT:    pcmpgtb %xmm8, %xmm12
>> +; SSE2-NEXT:    pcmpgtb %xmm3, %xmm12
>> ; SSE2-NEXT:    pcmpeqd %xmm0, %xmm0
>> ; SSE2-NEXT:    movdqa %xmm12, %xmm3
>> ; SSE2-NEXT:    pxor %xmm0, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm6, %xmm13
>> -; SSE2-NEXT:    pcmpgtb %xmm9, %xmm13
>> +; SSE2-NEXT:    pcmpgtb %xmm2, %xmm13
>> ; SSE2-NEXT:    movdqa %xmm13, %xmm2
>> ; SSE2-NEXT:    pxor %xmm0, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm5, %xmm14
>> @@ -4225,12 +4225,12 @@ define <16 x i32> @test114(<16 x i32> %a
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm8
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm9
>> -; SSE2-NEXT:    movdqa %xmm8, %xmm12
>> +; SSE2-NEXT:    movdqa %xmm3, %xmm12
>> ; SSE2-NEXT:    pcmpgtd %xmm7, %xmm12
>> ; SSE2-NEXT:    pcmpeqd %xmm13, %xmm13
>> ; SSE2-NEXT:    movdqa %xmm12, %xmm3
>> ; SSE2-NEXT:    pxor %xmm13, %xmm3
>> -; SSE2-NEXT:    movdqa %xmm9, %xmm14
>> +; SSE2-NEXT:    movdqa %xmm2, %xmm14
>> ; SSE2-NEXT:    pcmpgtd %xmm6, %xmm14
>> ; SSE2-NEXT:    movdqa %xmm14, %xmm2
>> ; SSE2-NEXT:    pxor %xmm13, %xmm2
>> @@ -4368,12 +4368,12 @@ define <16 x i32> @test116(<16 x i32> %a
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm9
>> ; SSE2-NEXT:    movdqa %xmm0, %xmm10
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm12
>> -; SSE2-NEXT:    pcmpgtd %xmm8, %xmm12
>> +; SSE2-NEXT:    pcmpgtd %xmm3, %xmm12
>> ; SSE2-NEXT:    pcmpeqd %xmm0, %xmm0
>> ; SSE2-NEXT:    movdqa %xmm12, %xmm3
>> ; SSE2-NEXT:    pxor %xmm0, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm6, %xmm13
>> -; SSE2-NEXT:    pcmpgtd %xmm9, %xmm13
>> +; SSE2-NEXT:    pcmpgtd %xmm2, %xmm13
>> ; SSE2-NEXT:    movdqa %xmm13, %xmm2
>> ; SSE2-NEXT:    pxor %xmm0, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm5, %xmm14
>> @@ -4890,7 +4890,7 @@ define <8 x i64> @test122(<8 x i64> %a,
>> ; SSE2-LABEL: test122:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm8
>> -; SSE2-NEXT:    movdqa %xmm8, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> +; SSE2-NEXT:    movdqa %xmm7, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm7
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> @@ -5164,7 +5164,7 @@ define <8 x i64> @test124(<8 x i64> %a,
>> ; SSE2-LABEL: test124:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm11
>> -; SSE2-NEXT:    movdqa %xmm11, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> +; SSE2-NEXT:    movdqa %xmm7, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm7
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> @@ -5467,7 +5467,7 @@ define <8 x i64> @test126(<8 x i64> %a,
>> ; SSE2-LABEL: test126:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm8
>> -; SSE2-NEXT:    movdqa %xmm8, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> +; SSE2-NEXT:    movdqa %xmm7, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm7
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> @@ -5795,7 +5795,7 @@ define <8 x i64> @test128(<8 x i64> %a,
>> ; SSE2-LABEL: test128:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm11
>> -; SSE2-NEXT:    movdqa %xmm11, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> +; SSE2-NEXT:    movdqa %xmm7, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm7
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> @@ -6047,7 +6047,7 @@ define <64 x i8> @test130(<64 x i8> %a,
>> ; SSE2-NEXT:    pcmpeqd %xmm13, %xmm13
>> ; SSE2-NEXT:    movdqa %xmm12, %xmm9
>> ; SSE2-NEXT:    pxor %xmm13, %xmm9
>> -; SSE2-NEXT:    movdqa %xmm8, %xmm14
>> +; SSE2-NEXT:    movdqa %xmm2, %xmm14
>> ; SSE2-NEXT:    pcmpgtb %xmm6, %xmm14
>> ; SSE2-NEXT:    movdqa %xmm14, %xmm2
>> ; SSE2-NEXT:    pxor %xmm13, %xmm2
>> @@ -6190,7 +6190,7 @@ define <64 x i8> @test132(<64 x i8> %a,
>> ; SSE2-NEXT:    movdqa %xmm12, %xmm9
>> ; SSE2-NEXT:    pxor %xmm0, %xmm9
>> ; SSE2-NEXT:    movdqa %xmm6, %xmm13
>> -; SSE2-NEXT:    pcmpgtb %xmm8, %xmm13
>> +; SSE2-NEXT:    pcmpgtb %xmm2, %xmm13
>> ; SSE2-NEXT:    movdqa %xmm13, %xmm2
>> ; SSE2-NEXT:    pxor %xmm0, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm5, %xmm14
>> @@ -6941,7 +6941,7 @@ define <16 x i32> @test146(<16 x i32> %a
>> ; SSE2-NEXT:    pcmpeqd %xmm13, %xmm13
>> ; SSE2-NEXT:    movdqa %xmm12, %xmm9
>> ; SSE2-NEXT:    pxor %xmm13, %xmm9
>> -; SSE2-NEXT:    movdqa %xmm8, %xmm14
>> +; SSE2-NEXT:    movdqa %xmm2, %xmm14
>> ; SSE2-NEXT:    pcmpgtd %xmm6, %xmm14
>> ; SSE2-NEXT:    movdqa %xmm14, %xmm2
>> ; SSE2-NEXT:    pxor %xmm13, %xmm2
>> @@ -7084,7 +7084,7 @@ define <16 x i32> @test148(<16 x i32> %a
>> ; SSE2-NEXT:    movdqa %xmm12, %xmm9
>> ; SSE2-NEXT:    pxor %xmm0, %xmm9
>> ; SSE2-NEXT:    movdqa %xmm6, %xmm13
>> -; SSE2-NEXT:    pcmpgtd %xmm8, %xmm13
>> +; SSE2-NEXT:    pcmpgtd %xmm2, %xmm13
>> ; SSE2-NEXT:    movdqa %xmm13, %xmm2
>> ; SSE2-NEXT:    pxor %xmm0, %xmm2
>> ; SSE2-NEXT:    movdqa %xmm5, %xmm14
>> @@ -7610,7 +7610,7 @@ define <8 x i64> @test154(<8 x i64> %a,
>> ; SSE2-LABEL: test154:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm8
>> -; SSE2-NEXT:    movdqa %xmm8, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> +; SSE2-NEXT:    movdqa %xmm7, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm7
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> @@ -7882,7 +7882,7 @@ define <8 x i64> @test156(<8 x i64> %a,
>> ; SSE2-LABEL: test156:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm11
>> -; SSE2-NEXT:    movdqa %xmm11, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> +; SSE2-NEXT:    movdqa %xmm7, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm7
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> @@ -8183,7 +8183,7 @@ define <8 x i64> @test158(<8 x i64> %a,
>> ; SSE2-LABEL: test158:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm8
>> -; SSE2-NEXT:    movdqa %xmm8, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> +; SSE2-NEXT:    movdqa %xmm7, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm7
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> @@ -8509,7 +8509,7 @@ define <8 x i64> @test160(<8 x i64> %a,
>> ; SSE2-LABEL: test160:
>> ; SSE2:       # BB#0: # %entry
>> ; SSE2-NEXT:    movdqa %xmm7, %xmm11
>> -; SSE2-NEXT:    movdqa %xmm11, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> +; SSE2-NEXT:    movdqa %xmm7, -{{[0-9]+}}(%rsp) # 16-byte Spill
>> ; SSE2-NEXT:    movdqa %xmm3, %xmm7
>> ; SSE2-NEXT:    movdqa %xmm2, %xmm3
>> ; SSE2-NEXT:    movdqa %xmm1, %xmm2
>> @@ -10289,7 +10289,7 @@ define <2 x i64> @test180(<2 x i64> %a,
>> ; SSE4:       # BB#0: # %entry
>> ; SSE4-NEXT:    movdqa %xmm0, %xmm2
>> ; SSE4-NEXT:    movdqa %xmm1, %xmm3
>> -; SSE4-NEXT:    pcmpgtq %xmm2, %xmm3
>> +; SSE4-NEXT:    pcmpgtq %xmm0, %xmm3
>> ; SSE4-NEXT:    pcmpeqd %xmm0, %xmm0
>> ; SSE4-NEXT:    pxor %xmm3, %xmm0
>> ; SSE4-NEXT:    blendvpd %xmm0, %xmm2, %xmm1
>> @@ -10768,7 +10768,7 @@ define <2 x i64> @test188(<2 x i64> %a,
>> ; SSE4:       # BB#0: # %entry
>> ; SSE4-NEXT:    movdqa %xmm0, %xmm2
>> ; SSE4-NEXT:    movdqa %xmm1, %xmm3
>> -; SSE4-NEXT:    pcmpgtq %xmm2, %xmm3
>> +; SSE4-NEXT:    pcmpgtq %xmm0, %xmm3
>> ; SSE4-NEXT:    pcmpeqd %xmm0, %xmm0
>> ; SSE4-NEXT:    pxor %xmm3, %xmm0
>> ; SSE4-NEXT:    blendvpd %xmm0, %xmm1, %xmm2
>>
>> Modified: llvm/trunk/test/CodeGen/X86/widen_conv-3.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/widen_conv-3.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/widen_conv-3.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/widen_conv-3.ll Mon Oct  2 15:01:37 2017
>> @@ -74,7 +74,7 @@ define void @convert_v3i8_to_v3f32(<3 x
>> ; X86-SSE2-NEXT:    cvtdq2ps %xmm0, %xmm0
>> ; X86-SSE2-NEXT:    movss %xmm0, (%eax)
>> ; X86-SSE2-NEXT:    movaps %xmm0, %xmm1
>> -; X86-SSE2-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1]
>> +; X86-SSE2-NEXT:    movhlps {{.*#+}} xmm1 = xmm0[1],xmm1[1]
>> ; X86-SSE2-NEXT:    movss %xmm1, 8(%eax)
>> ; X86-SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,2,3]
>> ; X86-SSE2-NEXT:    movss %xmm0, 4(%eax)
>>
>> Modified: llvm/trunk/test/CodeGen/X86/widen_conv-4.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/widen_conv-4.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/widen_conv-4.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/widen_conv-4.ll Mon Oct  2 15:01:37 2017
>> @@ -19,7 +19,7 @@ define void @convert_v7i16_v7f32(<7 x fl
>> ; X86-SSE2-NEXT:    movups %xmm0, (%eax)
>> ; X86-SSE2-NEXT:    movss %xmm2, 16(%eax)
>> ; X86-SSE2-NEXT:    movaps %xmm2, %xmm0
>> -; X86-SSE2-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
>> +; X86-SSE2-NEXT:    movhlps {{.*#+}} xmm0 = xmm2[1],xmm0[1]
>> ; X86-SSE2-NEXT:    movss %xmm0, 24(%eax)
>> ; X86-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1,2,3]
>> ; X86-SSE2-NEXT:    movss %xmm2, 20(%eax)
>> @@ -100,7 +100,7 @@ define void @convert_v3i8_to_v3f32(<3 x
>> ; X86-SSE2-NEXT:    cvtdq2ps %xmm0, %xmm0
>> ; X86-SSE2-NEXT:    movss %xmm0, (%eax)
>> ; X86-SSE2-NEXT:    movaps %xmm0, %xmm1
>> -; X86-SSE2-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1]
>> +; X86-SSE2-NEXT:    movhlps {{.*#+}} xmm1 = xmm0[1],xmm1[1]
>> ; X86-SSE2-NEXT:    movss %xmm1, 8(%eax)
>> ; X86-SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,2,3]
>> ; X86-SSE2-NEXT:    movss %xmm0, 4(%eax)
>>
>> Modified: llvm/trunk/test/CodeGen/X86/x86-interleaved-access.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/x86-interleaved-access.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/x86-interleaved-access.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/x86-interleaved-access.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -1721,7 +1721,7 @@ define void @interleaved_store_vf64_i8_s
>> ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = 
>> xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
>> ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm4 = 
>> xmm15[0],xmm5[0],xmm15[1],xmm5[1],xmm15[2],xmm5[2],xmm15[3],xmm5[3]
>> ; AVX1-NEXT:    vmovdqa %xmm8, %xmm1
>> -; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm11 = 
>> xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
>> +; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm11 = 
>> xmm8[0],xmm2[0],xmm8[1],xmm2[1],xmm8[2],xmm2[2],xmm8[3],xmm2[3]
>> ; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm11, %ymm0
>> ; AVX1-NEXT:    vmovups %ymm0, -{{[0-9]+}}(%rsp) # 32-byte Spill
>> ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm4 = 
>> xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
>>
>> Modified: llvm/trunk/test/CodeGen/X86/x86-shrink-wrap-unwind.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/x86-shrink-wrap-unwind.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/x86-shrink-wrap-unwind.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/x86-shrink-wrap-unwind.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -23,7 +23,7 @@ target triple = "x86_64-apple-macosx"
>> ; Compare the arguments and jump to exit.
>> ; After the prologue is set.
>> ; CHECK: movl %edi, [[ARG0CPY:%e[a-z]+]]
>> -; CHECK-NEXT: cmpl %esi, [[ARG0CPY]]
>> +; CHECK-NEXT: cmpl %esi, %edi
>> ; CHECK-NEXT: jge [[EXIT_LABEL:LBB[0-9_]+]]
>> ;
>> ; Store %a in the alloca.
>> @@ -69,7 +69,7 @@ attributes #0 = { "no-frame-pointer-elim
>> ; Compare the arguments and jump to exit.
>> ; After the prologue is set.
>> ; CHECK: movl %edi, [[ARG0CPY:%e[a-z]+]]
>> -; CHECK-NEXT: cmpl %esi, [[ARG0CPY]]
>> +; CHECK-NEXT: cmpl %esi, %edi
>> ; CHECK-NEXT: jge [[EXIT_LABEL:LBB[0-9_]+]]
>> ;
>> ; Prologue code.
>> @@ -115,7 +115,7 @@ attributes #1 = { "no-frame-pointer-elim
>> ; Compare the arguments and jump to exit.
>> ; After the prologue is set.
>> ; CHECK: movl %edi, [[ARG0CPY:%e[a-z]+]]
>> -; CHECK-NEXT: cmpl %esi, [[ARG0CPY]]
>> +; CHECK-NEXT: cmpl %esi, %edi
>> ; CHECK-NEXT: jge [[EXIT_LABEL:LBB[0-9_]+]]
>> ;
>> ; Prologue code.
>>
>> Modified: llvm/trunk/test/CodeGen/X86/x86-shrink-wrapping.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/x86-shrink-wrapping.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/x86-shrink-wrapping.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/x86-shrink-wrapping.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -17,7 +17,7 @@ target triple = "x86_64-apple-macosx"
>> ; Compare the arguments and jump to exit.
>> ; No prologue needed.
>> ; ENABLE: movl %edi, [[ARG0CPY:%e[a-z]+]]
>> -; ENABLE-NEXT: cmpl %esi, [[ARG0CPY]]
>> +; ENABLE-NEXT: cmpl %esi, %edi
>> ; ENABLE-NEXT: jge [[EXIT_LABEL:LBB[0-9_]+]]
>> ;
>> ; Prologue code.
>> @@ -27,7 +27,7 @@ target triple = "x86_64-apple-macosx"
>> ; Compare the arguments and jump to exit.
>> ; After the prologue is set.
>> ; DISABLE: movl %edi, [[ARG0CPY:%e[a-z]+]]
>> -; DISABLE-NEXT: cmpl %esi, [[ARG0CPY]]
>> +; DISABLE-NEXT: cmpl %esi, %edi
>> ; DISABLE-NEXT: jge [[EXIT_LABEL:LBB[0-9_]+]]
>> ;
>> ; Store %a in the alloca.
>>
>> Modified: llvm/trunk/test/DebugInfo/X86/live-debug-variables.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/DebugInfo/X86/live-debug-variables.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/DebugInfo/X86/live-debug-variables.ll (original)
>> +++ llvm/trunk/test/DebugInfo/X86/live-debug-variables.ll Mon Oct  2 
>> 15:01:37 2017
>> @@ -24,7 +24,7 @@
>>
>> ; CHECK:      .debug_loc contents:
>> ; CHECK-NEXT: 0x00000000:
>> -; CHECK-NEXT:              0x000000000000001f - 0x000000000000003c: 
>> DW_OP_reg3 RBX
>> +; CHECK-NEXT:              0x000000000000001f - 0x000000000000005a: 
>> DW_OP_reg3 RBX
>> ;   We should only have one entry
>> ; CHECK-NOT: :
>>
>>
>> Modified: llvm/trunk/test/DebugInfo/X86/spill-nospill.ll
>> URL: 
>> http://llvm.org/viewvc/llvm-project/llvm/trunk/test/DebugInfo/X86/spill-nospill.ll?rev=314729&r1=314728&r2=314729&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/DebugInfo/X86/spill-nospill.ll (original)
>> +++ llvm/trunk/test/DebugInfo/X86/spill-nospill.ll Mon Oct  2 15:01:37 
>> 2017
>> @@ -30,7 +30,7 @@
>> ; CHECK: callq   g
>> ; CHECK: movl    %eax, %[[CSR:[^ ]*]]
>> ; CHECK: #DEBUG_VALUE: f:y <- %ESI
>> -; CHECK: movl    %[[CSR]], %ecx
>> +; CHECK: movl    %eax, %ecx
>> ; CHECK: callq   g
>> ; CHECK: movl    %[[CSR]], %ecx
>> ; CHECK: callq   g
>>
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
> 

-- 
Geoff Berry
Employee of Qualcomm Datacenter Technologies, Inc.
  Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm 
Technologies, Inc.  Qualcomm Technologies, Inc. is a member of the Code 
Aurora Forum, a Linux Foundation Collaborative Project.


More information about the llvm-commits mailing list